inequality paper

THIN RATIONALITY REVIEW Gersen, Jacob; Vermeule, Adrian . Michigan Law Review ; Ann Arbor  Vol. 114, Iss. 8,  (Jun 2016): 1355-

1412.

ProQuest document link

ABSTRACT  

Under the Administrative Procedure Act, courts review and set aside agency action that is "arbitrary [and]

capricious." In a common formulation of rationality review, courts must either take a "hard look" at the rationality of

agency decisionmaking, or at least ensure that agencies themselves have taken a hard look. We will propose a

much less demanding and intrusive interpretation of rationality review-a thin version. Under a robust range of

conditions, rational agencies have good reason to decide in a manner that is inaccurate, nonrational, or arbitrary.

Although this claim is seemingly paradoxical or internally inconsistent, it simply rests on an appreciation of the

limits of reason, especially in administrative policymaking. Agency decisionmaking is nonideal decisionmaking;

what would be rational under ideal conditions is rarely a relevant question for agencies. Rather, agencies make

decisions under constraints of scarce time, information, and resources. Those constraints imply that agencies will

frequently have excellent reasons to depart from idealized first-order conceptions of administrative rationality.

Thin rationality review describes the law in action. Administrative law textbooks typically suggest that the State

Farm decision in 1983 inaugurated an era of stringent judicial review of agency decisionmaking for rationality.

That is flatly wrong at the level of the Supreme Court, where agencies have won no less than 92 percent of the

sixty-four arbitrariness challenges decided on the merits since the 1982 Term. The Court's precedent embodies an

approach to rationality review that is highly tolerant of the inescapable limits of agency rationality when making

decisions under uncertainty. State Farm is not representative of the law; beloved of law professors, and frequently

cited in rote fashion by judges, State Farm nonetheless lies well outside the mainstream of the Supreme Court's

precedent. To encapsulate the Court's approach to rationality review, the best choice would be the powerfully

deferential opinion in Baltimore Gas, decided in the same Term as State Farm. Plausibly, rather than living in the

era of hard look review or the State Farm era, we live in the era of Baltimore Gas. FULL TEXT  

Headnote

Under the Administrative Procedure Act, courts review and set aside agency action that is "arbitrary [and]

capricious." In a common formulation of rationality review, courts must either take a "hard look" at the rationality of

agency decisionmaking, or at least ensure that agencies themselves have taken a hard look. We will propose a

much less demanding and intrusive interpretation of rationality review-a thin version. Under a robust range of

conditions, rational agencies have good reason to decide in a manner that is inaccurate, nonrational, or arbitrary.

Although this claim is seemingly paradoxical or internally inconsistent, it simply rests on an appreciation of the

limits of reason, especially in administrative policymaking. Agency decisionmaking is nonideal decisionmaking;

what would be rational under ideal conditions is rarely a relevant question for agencies. Rather, agencies make

decisions under constraints of scarce time, information, and resources. Those constraints imply that agencies will

frequently have excellent reasons to depart from idealized first-order conceptions of administrative rationality.

Thin rationality review describes the law in action. Administrative law textbooks typically suggest that the State

Farm decision in 1983 inaugurated an era of stringent judicial review of agency decisionmaking for rationality.

That is flatly wrong at the level of the Supreme Court, where agencies have won no less than 92 percent of the

PDF GENERATED BY SEARCH.PROQUEST.COM Page 1 of 39

sixty-four arbitrariness challenges decided on the merits since the 1982 Term. The Court's precedent embodies an

approach to rationality review that is highly tolerant of the inescapable limits of agency rationality when making

decisions under uncertainty. State Farm is not representative of the law; beloved of law professors, and frequently

cited in rote fashion by judges, State Farm nonetheless lies well outside the mainstream of the Supreme Court's

precedent. To encapsulate the Court's approach to rationality review, the best choice would be the powerfully

deferential opinion in Baltimore Gas, decided in the same Term as State Farm. Plausibly, rather than living in the

era of hard look review or the State Farm era, we live in the era of Baltimore Gas.

Introduction

Under Section 706(2)(A) of the Administrative Procedure Act (APA), courts "shall" set aside agency action that is

"arbitrary [and] capricious."1 The conventional antonym of "arbitrary and capricious" is rational; as the D.C. Circuit

puts it, "[t]he 'arbitrary and capricious' standard deems the agency action presumptively valid provided the action

meets a minimum rationality standard."2 Hence, courts applying the arbitrary and capricious test review the

rationality of agency decisions.

The traditional and highly deferential approach, under the constitutional law of due process, equated rationality

review of agency decisionmaking with rational-basis review of legislation.3 Starting with Citizens to Preserve

Overton Park, Inc. v. Volpe4 in 1971, however, a vast and baroque caselaw elaborated the requirements of rational

agency decisionmaking under the APA. And in 1983, the Court in Motor Vehicle Manufacturers Association v. State

Farm Mutual Automobile Insurance Co. (State Farm)5 specifically held that rationality review of agencies under

706(2)(A) should be more demanding than rational-basis review.6 In a common formulation of rationality review,

courts must either take a "hard look" at the rationality of agency decisionmaking, or at least ensure that agencies

themselves have taken a hard look at the relevant problems. Hard look review is taken to encompass multiple

quasi-procedural obligations that, taken together, ensure agency rationality.7

We will propose a much less demanding and intrusive interpretation of the "arbitrary and capricious" standard in

section 706(2)(A). The argument has both prescriptive and descriptive components. Prescriptively, we urge that

rationality is a much thinner notion than some commentators seem to think, and that rational decisionmaking

requires far less from agencies than lawyers tend to realize.8 Courts have sometimes adopted an excessively

intrusive approach because, acting in the best of faith, they have misunderstood what rationality requires. In

particular, they have failed to grasp a crucial twist: under a robust range of conditions, rational agencies may have

good reason to decide in a manner that is inaccurate, nonrational, or arbitrary.9

Although this claim is seemingly paradoxical or internally inconsistent, it simply rests on an appreciation of the

limits of reason, especially in administrative policymaking. Agency decisionmaking is nonideal decisionmaking;

what would be rational under ideal conditions is rarely a relevant question for agencies. Rather, agencies make

decisions under constraints of scarce time, information, and resources. Those constraints imply that agencies will

frequently have excellent reasons to depart from idealized first-order conceptions of administrative rationality. We

will thus examine a series of limitations to agency rationality and to the communication of reasons by agencies,

and argue for an approach to rationality review that takes these limitations seriously-thin rationality review.

In a simplistic and idealized conception of administrative rationality, which is rarely articulated in explicit terms but

which implicitly underlies many judicial decisions, rational agencies should (1) attempt to choose the best policy

among the feasible options, after considering all relevant statutory factors and policy variables, and then (2)

explain to interested parties and to the court the agency's reasons for thinking that the chosen policy is best, as

compared to the alternatives.10 This conception turns out to be riddled with legal mistakes, conceptual slips, and

institutional problems.

As to (1), under the best reading of the arbitrary and capricious test, agencies have no legal obligation to consider

all policy variables that strike judges as arguably relevant. Agencies often have good reason to choose policies

that do not necessarily represent the best feasible option. Agencies may choose policies that the agency has not

compared to other feasible options. And agencies may choose policies that do not even produce net benefits in

the case at hand. The critical issue here is uncertainty, which sometimes gives agencies good second-order

PDF GENERATED BY SEARCH.PROQUEST.COM Page 2 of 39

reasons to depart from the simplistic firstorder conception of rationality under conditions we will identify.

As to (2), agencies may find it difficult to explain their reasons to generalist judges on reviewing courts, even, or

especially, when those reasons are valid. This is the problem of tacit expertise-tacit knowledge held by experts,

which is costly to transmit to nonexperts, and which is always distorted in the transmission. We do not suggest

that agencies should never be obliged to communicate reasons to courts, only that rationality review should

calibrate that obligation with sensitivity to the risk that genuine reasons are sometimes incommunicable between

experts and generalists, or at least costly to communicate. One of the main reasons agencies exist at all is

specialization, and specialization creates information asymmetries. Asymmetric information implies, in turn, that

the knowledge agencies possess cannot always be effectively evaluated by reviewing courts or other generalist

institutions, and sometimes it cannot even be effectively conveyed to those institutions.11

All this is negative critique, but our positive conception is straightforward. In contrast to thick rationality review,

the thin version posits that agencies are (merely) obliged to make decisions on the basis of reasons. Second-or-

higher order reasons may, in appropriate cases, satisfy that obligation. What is excluded by the arbitrary and

capricious standard is genuinely ungrounded agency decisionmaking, in the sense that the agency cannot justify

its action even as a response to the limits of reason. While truly capricious decisionmaking in this sense is no

doubt uncommon, it does exist; the caselaw contains examples.12 For purposes of interpreting section 706(2)(A),

"arbitrary and capricious" action is unreasoned agency action.

So much for prescription. Descriptively, our approach fits some but not all of the caselaw; it neither justifies all

extant decisions, nor condemns them all.13 Current law is actually a mixed bag-far more so than one might think

from reading administrative law textbooks, which typically suggest that State Farm inaugurated an era of stringent

judicial review of agency decisionmaking for rationality. As we will see, that suggestion is flatly wrong at the level

of the Supreme Court. At that level, agencies almost never lose. Indeed, the facts show that State Farm itself is an

outlier. Starting in October Term 1982, when State Farm was decided, the Court has passed on the merits of

arbitrariness challenges sixty-four times.14 Of those, agencies have lost arbitrary and capricious challenges only

five times-a remarkable win-rate of 92 percent.15

At the level of doctrine and announced principles, many of the modern cases feature strong rebukes of lower

courts for excessive interference, or specifically disavow idealized conceptions of rationality review developed by

lower courts. Over against State Farm there stands a long line of decidedly deferential decisions running from

Baltimore Gas &Electric Co. v. Natural Resources Defense Council, Inc.16 in 1983 to FCC v. Fox Television Stations,

Inc.17 and EPA v. EME Homer City Generation, L.P.18 in recent years. (The exceptions like Judulang v. Holder19

stand out, and are more easily recalled, precisely because they are rare.) This line of precedent embodies an

approach to rationality review that is more aware of, and tolerant of, the inescapable limits of rationality when

agencies make decisions under uncertainty-the chronic condition of decisionmaking in the administrative state.

State Farm is not representative of the law. Beloved by law professors and frequently cited in rote fashion by

judges, it nonetheless lies well outside the mainstream of the Supreme Court's precedent.

Of course there are many possible explanations for these data, which we will examine in detail. Selection effects

must be accounted for, both at the stage of the selection of cases for litigation, and at the stage of selection of

cases for review by the Supreme Court. Agencies' won-loss record in arbitrariness challenges before the D.C.

Circuit is somewhat less impressive, although the data is sketchy. (As we will see, there may well be a connection

between the D.C. Circuit's episodic bullying of agencies and the Supreme Court's marked tendency to validate

agency decisions in the face of legal challenges.) At a minimum, however, administrative lawyers need to reassess,

perhaps dramatically, the folk wisdom about the era of hard look review. Forced to pick one case to encapsulate

the Court's approach to rationality review, the best choice would be the powerfully deferential opinion in Baltimore

Gas, decided in the same term as State Farm. Plausibly, rather than living in the era of hard look review or the State

Farm era, we live in the era of Baltimore Gas.

In Baltimore Gas, decided the same term as State Farm, the Supreme Court upheld the Nuclear Regulatory

Commission's determination that, for purposes of licensing under the National Environmental Policy Act, the

PDF GENERATED BY SEARCH.PROQUEST.COM Page 3 of 39

permanent storage of certain nuclear wastes would be assumed to have no significant environmental impact (the

"zero-release" assumption).20 Per Judge Bazelon, the D.C. Circuit had struck down the agency's judgment on

arbitrariness grounds.21 The Court reversed that decision in emphatic terms. Quoting Vermont Yankee, Justice

O'Connor wrote that "[a]dministrative decisions should be set aside in this context, as in every other, only for

substantial procedural or substantive reasons as mandated by statute . . . not simply because the court is unhappy

with the result reached."22 Moreover, "a reviewing court must remember that the Commission is making

predictions, within its area of special expertise, at the frontiers of science. When examining this kind of scientific

determination, as opposed to simple findings of fact, a reviewing court must generally be at its most

deferential."23 In our view, it is time to stop discussing State Farm whenever arbitrariness review is mentioned.

The better reflex-better theoretically, and more faithful to the caselaw-would be for lawyers to mention "Baltimore

Gas" review.

Part I lays out the descriptive claim. We begin by distinguishing two versions of State Farm-the case itself on the

one hand, and on the other, the broader culture of hard look review for which the case has become a symbol and

shorthand. We then state some basic findings about the Court's caselaw since State Farm, consider various

explanations for those findings, and underscore their inconsistency with a simple narrative about the prevalence of

hard look review. Although there are many possible explanations for the facts, it is clearly false that the Court is in

any sense committed to stringent supervision of agency rationality.

In Part II, the prescriptive section, we discuss the limitations of agency rationality, especially when making

decisions under uncertainty, and the limitations on agency ability to communicate reasons to generalist reviewers.

Our main claim is that administrative law should, and often does, recognize that agencies have good second-order

reasons for departing from first-order rationality. Accordingly, so long as agencies act on the basis of reasons,

including second- or higher-order reasons, the thin conception of rationality review is satisfied. We modulate this

claim by examining similarities and differences between different types of reviewers-most notably generalist

judges, on the one hand, and the reviewers of a different type (a mix of economists and lawyers) who staff the

Office of Information and Regulatory Affairs (OIRA) on the other. Although some of our points apply to OIRA as

well as to the federal judiciary, some do not. In any event, OIRA is not central for our project. Judicial review is our

focus, and hard look review by judges is our principal target.

Taken together, the descriptive and prescriptive sections offer a unified account of rationality review. The caselaw

and the theory march in tandem to a surprising extent, although with notable exceptions and outliers. To some

substantial degree, judicial practice already embodies the thin approach to rationality review that theory

recommends-an approach that is far more flexible, accommodating, and intelligent about agency rationality and its

inescapable limitations than is the hard look approach.

I. Description: The Real World of State Farm

A. State Farm and "State Farm"

A problem with our topic is that there are really two versions of State Farm. One is the opinion itself, which is

narrower than many commentators have made it out to be. Another is the broader aura of the decision.

Commentators, generally suspicious of agency rationality, have puffed up State Farm into a synecdoche for hard

look review, contributing to a pervasive but latent culture of academic skepticism towards agency explanations

and agency decisionmaking. Just as there are two versions of Marbury v. Madison-the rather narrow decision

itself, and the inflated and heavily symbolic Marbury v. Madison that the Court so routinely invokes as a symbol of

judicial supremacy24-so too State Farm as doctrine coexists uneasily with "State Farm" as symbol.

Our target is the latter, not the former. State Farm itself, we will argue, is in important respects less demanding

than the lawyers' culture of hard look suggests. The decision is clear, for example, that judges have no warrant to

require agencies to consider and discuss any policy alternative that the judges happen to believe is relevant to the

problem at hand.25 And we will see that the Supreme Court has been careful, over time, about policing the

boundaries and maintaining the limits of State Farm. But the problem is that lower courts and (especially)

commentators have sometimes been less careful. Overreading State Farm, they have applied it as though it

PDF GENERATED BY SEARCH.PROQUEST.COM Page 4 of 39

demands a kind of unbounded, ideal rationality from agencies, insensitive to the costs of information and

decisionmaking. Thin rationality review, by contrast, emphasizes that agencies are constrained by limited

resources, information, and time, and asks what (nonideal) reasons agencies may have for acting inaccurately,

nonrationally, or arbitrarily in light of those limits.

While "State Farm" as symbol is not the law, there is nothing in State Farm itself that is incompatible with our

approach. Suppose judges applying State Farm started to fold in all the second-order reasons and nonideal

constraints we will discuss, and they started to say that even after a "hard look," agencies may act inaccurately,

nonrationally, or arbitrarily, where agencies have good reason to do so. Then we would have no quarrel with the

resulting caselaw, but only because our normative claim-that good decisionmaking allows for departures from

first-order rationality where there is adequate reason for such departures-would have been built into hard look

review itself. (In which case, of course, "hard look" would then be something of a misnomer).

B. Some Facts About the Law

What does the real world of arbitrary and capricious review look like, and where do we look to find out? Our

descriptive survey is impressionistic, but suggestive. We begin with an overview of State Farm in the Supreme

Court and then turn to the courts of appeals. There is a distribution of cases in both settings with important cases

in the tails. Nevertheless, the outliers should not be allowed to blur the overall picture. There are, of course,

instances of aggressive and intensive review under the arbitrary and capricious framework; but in the Supreme

Court at least, agencies almost always win. If the task for appellate courts is to practice what the Supreme Court

both preaches and practices, the message is to apply a thin form of rationality review. In the run of cases, arbitrary

and capricious review entails a predictably and sensibly deferential review of agency policy judgments.

1. The Supreme Court

Since the 1982-1983 Term, when State Farm was decided, the Court has passed on the merits of arbitrariness

challenges sixty-four times.26 The agency lost on arbitrariness grounds in only five cases, and in three additional

cases the agency lost on statutory and arbitrariness grounds.27 What should we make of the fact that the Court

deems agency decisions arbitrary at most 13 percent of the time, and indeed only 8 percent of the time if we

confine ourselves to pure arbitrariness cases? Clearing the hard look hurdle in the Supreme Court is hardly a heroic

task.

To make headway on this question, consider first the hard cases for our proposition: those in which the agency's

decision was struck down as arbitrary. To be sure, these cases are outliers, but like all exceptions, they are

important data points for understanding the rule. Two of these cases, Massachusetts v. EPA28 and Morgan

Stanley Capital Group, Inc. v. Public Utility District No. 1,29 actually involve agency decisions not to act at all.

Famously, Massachusetts v. EPA overturned EPA's decision not to decide whether greenhouse gases constituted

"pollution" pursuant to the Clean Air Act.30 But this was hardly a run-of-the-mill hard look case. While agency

decisions not to act-at least decisions not to engage in rulemaking-may be subject to hard look review,31 in

practice, agency failures to act tend to reach the Supreme Court in a posture of near-complete failure by the

agency to carry out its statutory obligations on a given issue.32 When the agency loses in such cases, the Court is

saying, in effect, that the agency must do something, but that is a far cry from intensive hard look review of an

ultimate agency decision. When the agency has decided which of several policy options to pursue, the Supreme

Court almost never strikes down that judgment as arbitrary. It is when the agency has said "we will do nothing"

that the Supreme Court is willing to step in.

The Supreme Court case that most directly challenges our view is probably Judulang v. Holder.33 The case

involved a challenge to the Board of Immigration Appeals' (BIA) policy for deciding when resident aliens may apply

to the Attorney General for relief from deportation. Justice Kagan's opinion observed that "[w]hen an administrative

agency sets policy, it must provide a reasoned explanation for its action. That is not a high bar, but it is an

unwavering one."34 The key language here is "not a high bar." In concluding that the BIA's "comparable grounds"

approach to determining immigration status was entirely arbitrary, the Court explained that the decision rule must

bear some relationship to relevant statutory facts or factors.35 The Court explained that the agency's policy turned

PDF GENERATED BY SEARCH.PROQUEST.COM Page 5 of 39

on issues having nothing to do with even arguably relevant facts: "Rather than considering factors that might be

thought germane to the deportation decision, that policy hinges §212(c) eligibility on an irrelevant comparison

between statutory provisions."36 As important, the Court made clear that the factors that must be considered are

not limited to those expressly identified in the statute. Rather, the Court said that legitimate factors for the agency

to consider for arbitrariness review are those "tied, even if loosely, to the purposes of the immigration laws or the

appropriate operation of the immigration system."37 The BIA lost in the case because it lacked even this level of

tangential support for its policy, but Judulang v. Holder actually clarifies that agencies have more latitude to select

factors to explain or justify their decisions than has traditionally been taught. It seems clear that this most recent

instance of hard look review is remarkably modest and far more consistent with Baltimore Gas than State Farm

itself. And more broadly, the few cases in which the agency loses before the Court feature either refusals to act

altogether, or the adoption of a decision rule that is entirely unrelated to any statutorily relevant factor, and in that

sense genuinely capricious.

Our main point, however, is that the rare cases like Judulang v. Holder must not be allowed to take up too much

space in the collective memory of the profession. To a first rough approximation, agencies almost always win

arbitrariness challenges in the Supreme Court.38 And to be clear, the distribution of these cases is diverse. Six of

the cases involve challenges to Environmental Protection Agency actions, including enforcement orders, permit

decisions, and notice and comment rules.39 Two cases review Board of Immigration proceedings.40 Five cases

review decisions of the Federal Communications Commission, both rate-setting, enforcement actions, and several

informal rulemakings: no action was held to be arbitrary and capricious.41 The agency represented most in the

sample of Supreme Court challenges is the Department of Health and Human Services (HHS), which saw fourteen

challenges before the Court. In all but one, the agency's action was upheld as not arbitrary and capricious. The

other cases involved a smorgasbord of agencies, including the National Labor Relations Board (three enforcement

orders upheld as not arbitrary and capricious), the Department of the Interior, the Department of Commerce, the

Securities and Exchange Commission, and the Merit Systems Protection Board. The agencies at issue are quite

heterogeneous, and the cases include a mix of actions from informal rulemaking, formal adjudication, enforcement

actions, licensing and permit decisions, to rate-setting, and so on. In case after case, no matter the agency and no

matter the action, the most likely outcome by an overwhelming margin is for the Court to uphold the action as not

arbitrary and capricious. Nor does there seem to be much partisan disagreement about the thinness of review. The

cases upholding agency action include majority opinions by Justices Scalia, Rehnquist, Roberts, Souter, Stevens,

Ginsburg, Thomas, White, Breyer, Powell, Kennedy, Blackmun, Alito, Kagan, and O'Connor. In short, all of the

justices are well represented in the thin rationality review camp.

2. Courts of Appeals

In the courts of appeals, the record of arbitrary and capricious review is more mixed. We consider both large-N

empirical evidence and then an impressionistic survey of some recent decisions. There are quite a few empirical

studies of judicial review of agency action-broadly defined-in the courts of appeals, but few of arbitrary and

capricious review specifically. Miles and Sunstein remarked in 2008 at the "sparse empirical literature . . . on the

actual operation of the hard look doctrine," noting additionally that "[t]here is no systematic evidence on the rate of

invalidation under hard look review."42 In reviewing empirical studies across multiple judicial review doctrines,

Richard J. Pierce, Jr. noted in 2011 that only Miles and Sunstein had examined agency success rates in the courts

of appeals under State Farm.43

Miles and Sunstein conclude that agencies win 64 percent of court of appeals cases reviewing their decisions for

arbitrariness.44 Their data set includes "all published appellate rulings from 1996 to 2006 involving review of

decisions of the EPA and review of NLRB decisions either for arbitrariness or for lack of substantial evidence."45

The authors sensibly limited their analysis to decisions reviewing those two agencies because the issues at stake

mapped easily onto political worldviews-the primary focus of the article. But if one looks at the distribution of

actual cases, just under 85 percent of the cases they study are NLRB cases.46 The NLRB is an important agency,

but as students of administrative law will quickly recognize, the board makes policy in a way that is almost unique

PDF GENERATED BY SEARCH.PROQUEST.COM Page 6 of 39

in the universe of administrative agencies, proceeding to formulate rules of general applicability through

adjudication rather than rulemaking.47 Moreover, as the authors note, the NLRB's decisions-when it does anything

at all-tend to be more consistently "liberal" by traditional metrics, making them targets of whatever conservative

judicial politics might exist.48

Thus, the 64 percent figure is somewhat misleading. Judges voted to overturn far more consequential EPA

decisions on arbitrariness grounds in barely more than one-fourth of the cases.49 EPA decisions are upheld

against arbitrariness challenges nearly 80 percent of the time.50 This is, of course, a partial sample of the arbitrary

and capricious universe-NLRB and EPA decisions during 1996-2006-but there is little in the data to suggest that

hard look is being used as an elaborate or onerous form of review.

More recently, however, there have been several notable cases in which courts of appeals have adopted a form of

hard look review that is consistent with neither the Supreme Court's principles and practice, nor even with past

lower court approaches. In Business Roundtable v. SEC, the D.C. Circuit struck down SEC rule 14a-11, which

required "public companies to provide shareholders with information about, and their ability to vote for,

shareholder-nominated candidates for the board of directors."51 The court concluded that the SEC's Rule was

arbitrary and capricious: "Because the agency failed to 'make tough choices about which of the competing

estimates is most plausible, [or] to hazard a guess as to which is correct,' we believe it neglected its statutory

obligation to assess the economic consequences of its rule."52 So far, the reaction to Business Roundtable has

been mixed, but the modal response among administrative lawyers has been a mix of surprise and dismay.53

Importantly, some have taken the case to stand for the proposition that an agency rule is arbitrary and capricious if

it is not supported by careful and rigorous cost-benefit analysis, including a detailed statement of any potential

costs or benefits that cannot be quantified and a clear statement about how and why competing estimates of

costs were resolved.54 So interpreted, the case stands for an ambitious form of arbitrariness review that requires

cost-benefit analysis to the extent possible, unless statutorily precluded.55

On this view, Business Roundtable constitutes an outlier approach to arbitrariness review, in which the court

required an elaborate showing of costbenefit analysis where possible. Shortly we will argue in general terms that

this approach neither is, nor should be, the law. As for the case, suffice it to say that Business Roundtable

illustrates the problems of hard look judicial review, particularly with respect to economic market regulation. Many

of the potential effects, and therefore the costs and benefits, of the rule would have been a function of a complex

set of interdependent strategies and replies by firms, investors, competitors, and so on-a form of strategic

uncertainty.56 Both the benefits and the costs of the rule, and others like it, are uncertain precisely because it is

difficult to predict with any precision how markets will react.57 No doubt there are those who believe that this

means the SEC should have stayed its hand, but there is a good deal of space between that view and a conclusion

that the SEC may not lawfully enact a rule requiring notice about ways in which shareholders may nominate and

elect directors. We certainly have no view on the wisdom of Rule 14a-11. But in a domain where the consequences

of the rule are uncertain, the stakes high, and the importance of specialized knowledge critical, there is no reason

to think that a three-judge panel knows better than the SEC. We will return to these themes in the next Part.

All that said, let us keep the larger picture firmly in view. Unusual cases tend to grip the mind. But Business

Roundtable, and a set of other related or similar cases,58 are outliers. The days of systematically aggressive hard

look review, as in the D.C. Circuit's decisions from the 1970s and early 1980s, are mostly behind us, thanks in part

to the Supreme Court, which sat down heavily on the D.C. Circuit in both Vermont Yankee59 and Baltimore Gas.60

And the Court recently overturned another procedural innovation from the D.C. Circuit in Perez v. Mortgage

Bankers Association.61

3. Problems

Judicial decisions are the result of a complex set of anticipated behaviors by agencies and courts. If agencies

believe that judicial review will be aggressive, then strategic agencies may be more conservative in their policy

choices, adopting decisions that are well justified by the available evidence. As a consequence, agencies would

usually win in litigation, making it appear that courts are applying a highly deferential standard. By the same token,

PDF GENERATED BY SEARCH.PROQUEST.COM Page 7 of 39

if agencies believe judicial review will be modest, they might be more aggressive in their decisions, stretching

policies just to the edge of justifiability and beyond. If so, agencies might lose a lot in litigation, making it appear

that courts are adopting a stringent form of review. These are familiar dynamics that travel under the rubric of

selection effects.62

Selection effects are not a challenge to our prescriptive vision of arbitrariness review, but they are a potential

problem for our descriptive account of arbitrariness review. To illustrate, we have claimed that agencies virtually

never lose on arbitrariness grounds in the Supreme Court, and that agencies rarely lose on arbitrariness grounds in

the courts of appeals. We have interpreted that empirical regularity to suggest that the Supreme Court is directing

the lower courts to utilize a thinner form of rationality review, one that requires merely that the agency's decision

not be pure caprice. When the Supreme Court takes an arbitrariness case on review, it is almost always to support

the agency's decision. Selection effects, however, threaten our interpretation, or at least throw a fly into the soup.

Suppose we are correct and the courts apply a form of thin rationality review. If agencies recognize this, they might

begin to relax the rigor with which they justify their own decisionmaking. Courts, applying the same standard of

thin rationality review, would suddenly strike down more agency actions, not because the court has adopted more

aggressive review, but because the agency has weakened its decisionmaking. Similarly, suppose the courts are

engaging in an ambitious thick form of rationality review. Agencies, recognizing this practice, may begin to beef up

their decisionmaking process and justify their conclusions with more elaborate and careful consideration. As a

result, courts-again, applying the same thick form of rationality review-would uphold many more agency decisions,

making it appear that there was a new easy bar to clear over which agencies virtually never trip. Furthermore, when

agencies lose in the courts of appeals, the Solicitor General (SG) often acts as a gatekeeper whose consent

agencies must win in order to file a certiorari petition in the Supreme Court. If the SG files petitions only when the

agency's case is strong, agencies might have a better win-rate in the Supreme Court than in the lower courts, as

indeed they do. All these possibilities are intrinsically speculative. We cannot say anything definitive about

selection effects, nor can proponents of hard look review. We can, however, be intellectually candid about the

selection-effects problem, a practice we hope will be embraced by others as well.

That said, there is some indirect evidence we can bring to bear. In what is perhaps (we hope) the definitive word on

empirical studies of agency winrates in litigation, David Zaring performed a meta-analysis of existing data and also

added new data on judicial review of agency fact-finding.63 Zaring found that no matter what standard of review

was utilized (hard look, Chevron, or substantial evidence), and no matter what aspect of the decision was reviewed

(policy, law, or fact), agency win-rates were surprisingly stable.64 Almost uniformly, agencies won in litigation

about 70 percent of the time.65 Neither the standard of review nor the aspect of the underlying decision being

challenged seemed to matter much at all. To be sure, selection effects might be operating offstage in all these

settings, including in the lower courts, but at a certain point that abstract possibility ceases to impress. The

important point for our purposes is that no matter what linguistic formulation courts invoke when engaging in

rationality review, courts don't seem to be engaging in hard look analysis-or at least it requires major epicycles

about selection effects to save the hard look story. Agencies usually win, and rather than requiring anything like a

searching hard look inquiry, the resulting distribution of agency wins is far more consistent with a thin form of

rationality review.

Another type of indirect evidence involves the putative problem of "ossification." The myth of rigorous State Farm

review has long been accompanied by several mini-myths, part of the inherited generational wisdom about the

costs of intensive judicial review of agency policymaking. Chief among these is the problem of agency

ossification: consistent agency losses in litigation, coupled with fear of judicial review, resulted in agency

paralysis. Informal rulemaking began taking years, if not decades; rulemaking records grew exponentially as

agencies were forced to address every possible concern, with respect to every possible issue in every rulemaking,

no matter how large or small.66 The State Farm version of hard look review made it increasingly difficult for

agencies to do their jobs-or so we have been taught.

If true, agency ossification would be some empirical evidence in favor of the strong version of State Farm review.

PDF GENERATED BY SEARCH.PROQUEST.COM Page 8 of 39

Like rigorous State Farm review itself, however, the ossification phenomenon is long on anecdote and short on

data. In fact, recent studies of agency rulemaking find virtually no evidence of the ossification thesis. EPA rules go

from start to finish in an average of a year and a half.67 In one exhaustive study of the unified regulatory agenda,

Anne O'Connell concludes that the "costs to rulemaking . . . are certainly not so high as to prohibit considerable

rulemaking activity."68 And, in another comprehensive survey of agency regulatory activity, Yackee and Yackee

conclude that agencies are able to promulgate large numbers of rules fairly quickly.69

If there is little evidence of actual ossification, why has the idea of ossification had so much influence? In part, it is

because the few outlier examples are high profile and therefore highly visible; but for that very reason, one cannot

accurately generalize from those cases to the remainder of the distribution of rulemakings. We suspect, however,

that there is a simpler explanation as well. If one believes that State Farm entails searching hard look review, then

it simply stands to reason, as a matter of nearly inimitable logic, that ossification will result. Once that misguided

assumption is relaxed, the fact that there is little evidence of ossification makes perfect sense. Arbitrariness

review is like a legal phantom: it can scare, but rarely hurts. So long as agencies comply with some minimal

rationality requirements, they usually win in litigation.

The possibility of strong selection effects complicates our descriptive claim about agencies' overwhelming win-

loss record in the Supreme Court and, a fortiori, their impressive-but not overwhelming-win-loss record in the

courts of appeals. However, the selection effects hypothesis is speculative, and there is indirect evidence that it is

at most a weak force-evidence both from Zaring's global comparison of standards of review and from the

ossification studies. And let us be clear that what the Court says is even more important than the win-loss record.

As we will detail in the next Part, the Court's approach to rationality review is thin in precisely our sense. Justices

of all types-unlike some lower court judges-show a deep appreciation for the constraints under which agencies act,

high tolerance for agency action under uncertainty, and a willingness to allow agencies to adopt strategies of

second-order rationality that permit inaccurate, nonrational, or arbitrary action in particular cases.

II. Prescription: Thin Rationality Review

What if anything might justify the pattern of deferential rationality review we have observed in Part I? There is a

theory implicit, and sometimes explicit, in the bulk of the Court's caselaw: thin rationality review. In this Part, we lay

out the theory in both negative and positive terms, and examine its institutional implications for judicial review of

agency action. We will also examine some indirect implications for nonjudicial review of agency action, OIRA

review being the main example.

We begin, in Section II.A, with a series of negative claims. There are many things that rationality review does not

require, and that the Court has generally disavowed, despite contrary assumptions or arguments scattered through

the lower court caselaw and especially the commentary. Judges may not require agencies to conduct quantified

cost-benefit analysis, even presumptively; may not always require agencies to conduct comparative policy

evaluation, obliging agencies to show that the chosen policy is superior to feasible alternatives, or superior to the

agency's own past choices; may not require agencies to have valid first-order reasons for all their choices; may not

force agencies to opt for "conservative" assumptions in the face of uncertainty; need not require a "rational

connection between the facts found and the choices made,"70 depending upon how exactly that critical idea is

understood; need not require agencies to be able to explain or convey their reasons, to the satisfaction of a panel

of generalist judges; and may not lard rationality review with quasi-procedural obligations.

In Section II.B, we offer a simple positive formulation of rationality review that attempts to capture the thin

conception. We hold that agencies must act based on reasons, but also that (1) agencies may have reasons they

cannot give, at least at acceptable cost; and (2) the set of admissible reasons includes second-order reasons to

act inaccurately, nonrationally, or arbitrarily. That formulation seems paradoxical; we will attempt to explain that

the paradox is illusory.

Finally, in Section II.C, we will consider whether and how our thin conception of rationality review applies, not only

to judicial review, but to review by OIRA or other executive branch institutions. Some, but not all, of our negative

prescriptions hold, mutatis mutandis-most controversially, our suggestions that reviewers have no basis in the

PDF GENERATED BY SEARCH.PROQUEST.COM Page 9 of 39

theory of rational decisionmaking for requiring agencies to conduct quantified cost-benefit analysis (although

there may be other grounds for doing so), and that agencies may have good reasons that they may be unable to

communicate to the reviewer. Most generally, our claim that agencies may have good secondorder reasons to act

in ways that are inaccurate, nonrational, or arbitrary holds, regardless of the nature of the reviewing body.

A. Negative Prescriptions

1. "Relevant Factors"

To clear some ground, we will briefly mention a recurring confusion that stems not from State Farm, but from the

original framework for rationality review laid down in Overton Park. Chief among the latter's innovations was the

idea that agencies must consider "the relevant factors," and also avoid "clear error[s] of judgment."71 The

injunction to consider the "relevant factors" has fostered nontrivial confusion, because the source of those factors

is not obvious. Is Overton Park saying, for example, that judges should identify policy factors in the problem at

issue, policy factors that seem relevant to them, and then require agencies to consider those factors?

Not at all; indeed, the Court has specifically repudiated that procedure both in State Farm itself72 and in

subsequent cases.73 Understood in the larger framework of judicial review of agency action, the function of the

"relevant factors" inquiry is simply to ensure that the agency has given due consideration to any factors made

relevant by the authorizing statute itself, and to ensure that the agency has not considered any factors the statute

rules off-limits. Absent constitutional problems, Congress enjoys the power to set the agency's deliberative agenda

by statute, either in positive or negative terms, and the "relevant factors" inquiry ensures that agencies respect

Congress's choices.

Given that the relevant factors inquiry is really one of statutory interpretation, it is subject to the rules of statutory

interpretation that always govern in administrative law. One of those is the Chevron doctrine, under which

agencies, rather than courts, enjoy the authority to fill in statutory gaps and ambiguities. The Court has made it

plain that Chevron applies to the interpretive question about what factors the statute makes relevant.74 And, three

terms ago, the Court also explained that Chevron applies to agency interpretations of their own jurisdiction as

well.75 In particular, where statutes are silent or ambiguous, agencies-rather than courts-enjoy discretion to decide

what the relevant factors may be and whether to consider those factors.76

Agencies also enjoy discretion to decide when to consider those factors. The Court has also been clear that

agencies need not consider all logically relevant policy factors at once, but may instead parcel them out into

different proceedings, considering problems by parts and proceeding one step at a time.77 Relatedly, agencies

may in adjudication single out one or a few defendants from the mass of similarly situated firms, in order to create

a test case, or to examine the relevant questions case by case-even if from the defendant's point of view, the

selection is entirely arbitrary.78

This relaxed approach makes eminent sense. The precondition for Chevron to apply is a "step zero" analysis, which

asks whether Congress has delegated law-interpreting authority to agencies rather than courts.79 The animating

objectives of Chevron-political accountability and expertise- both suggest that Congress's default intention is that

agencies, rather than courts, should determine which policy factors count as "relevant," and which factors should

be considered in which proceeding, absent clear statutory indication to the contrary or some special reason to

think that the question is so important that it is not fit for agency resolution.80 As to any reasonably complex

policy problem, an indefinitely large number of policy factors are potentially relevant, or can be claimed to be

relevant by litigants who benefit from delaying agency action. Generalist judges who attempt to sift the wheat

from the chaff will run every risk of becoming confused, absent explicit statutory guidance, and will inevitably end

up making de facto policy choices that should lie within the province of relatively more responsive and better-

informed agencies.

So an agency that otherwise enjoys delegated interpretive authority under Chevron "step zero" also enjoys the

authority to decide which factors count as "relevant" for purposes of Overton Park-provided, of course, that the

underlying statute is silent or ambiguous. Where statutes are clear, however, either mandating or prohibiting

consideration of relevant factors, courts must enforce their terms. In other words, the ordinary Chevron inquiry

PDF GENERATED BY SEARCH.PROQUEST.COM Page 10 of 39

governs. For our purposes, the important point is that rationality review neither requires, nor even permits,

generalist judges to decide on their own initiative that a given factor that happens to strike them as important is a

legally "relevant factor" under Overton Park . Absent clear statutory instruction or an issue of extraordinary political

and economic significance, policy relevance is a matter for agency determination.

2. Cost-Benefit Analysis-Quantified and Otherwise

With the advent of Business Roundtable, some have begun to suggest that cost-benefit analysis is a necessary

component of rational decisionmaking, so that a requirement of cost-benefit analysis should be read into arbitrary

and capricious review.81 How can a decision be rational if it is unjustified by the attendant costs and benefits?82

Legally, there are two related-but-distinct ideas. As to the interpretation of agency organic statutes, one view holds

that congressional silence or ambiguity should be read to require cost-benefit analysis, quantified where

possible.83 In a related view, unless statutes clearly prohibit the consideration of costs, arbitrariness review

should be understood to require that agencies supply quantified cost-benefit analysis, wherever quantification is

possible.84

In our view, however, these claims hover between confused and mistaken. There is a thin tautological sense in

which rationality requires that decisionmakers do what is better, as opposed to what is worse. But rationality

certainly does not require quantified cost-benefit analysis in the technical sense. As for the interpretation of

organic statutes and of arbitrariness review under the APA, there is no plausible basis for even a presumptive

requirement of quantified cost-benefit analysis.

a. Conceptual Problems

At the conceptual level, there is slippage in this literature between a tautology on the one hand, and a highly

sectarian decision-procedure on the other.85 The tautology is that a decisionmaker should do what is best, all

things considered. In that sense, it may always be said, without possibility of contradiction, that the decisionmaker

should ensure that the "benefits" exceed the "costs." Were a decisionmaker to say that "the benefits are X; the

costs are Y; we find that Y >X; nonetheless we choose to do X," one would either doubt the decisionmaker's

rationality or assume some misunderstanding or miscommunication.86 (As we will see, however, this sort of

example assumes certainty, or at least well-formed probability assessments, as to the costs and benefits. Under

conditions of genuine uncertainty, perhaps arising from high costs of information, rationality even permits

decisionprocedures that do not attempt to figure out, in a first-order way, whether benefits exceed costs.)

Thus an informal and nonquantified sense of cost-benefit analysis- thinking about the "pros" and "cons"87-is

ubiquitous, both in law and life. Charles Darwin famously drew up a list of pros and cons in deciding whether to

marry.88 But this informal sense of cost-benefit analysis is not what is at stake in the legal debates. In Business

Roundtable, the SEC offered a detailed qualitative discussion of the pros and cons of its rule.89 The court

nonetheless objected that the agency's cost-benefit analysis was inadequate because the agency had not tried

hard enough to quantify some of the relevant factors.90

As Business Roundtable illustrates, proponents urge a particular, highly structured decision-procedure: quantified

CBA.91 That procedure is technical, but it is also highly controversial, even polarizing, especially in its purest form,

which is not only quantified but monetized. Proponents praise it as a mechanism for welfare-maximization, for

promoting democratic transparency in agency decisionmaking, or for securing presidential control of agencies.92

Opponents criticize it for reducing incommensurables to a common denominator, for smuggling in controversial

value judgments and hidden margins of discretion, and for crowding out nonquantifiable considerations.93

The problems and promise of quantified CBA have been rehashed many times, and we do not address its merits

here. But quantified CBA is a specialized, sectarian decision-procedure, not a requirement of rational

decisionmaking. Many demonstrably rational economists, philosophers, lawyers and other students of

decisionmaking do not believe that rationality requires quantified CBA. Some believe, on the contrary, that best

practices of decisionmaking actually forbid resort to quantified CBA, because it has no moral relevance

whatsoever and sometimes misleads.94 More temperately, others believe that quantified CBA is sometimes

useful, sometimes not, but reject the idea that it is inscribed in the very nature of rationality.95 Quantified CBA is

PDF GENERATED BY SEARCH.PROQUEST.COM Page 11 of 39

both disputable and widely disputed. To impose it on agencies in the name of rationality would be to squelch

reasonable disagreement by sheer force.

b. Legal Problems

The legal issue is whether judges may require agencies to use quantified CBA, at least presumptively, either as a

matter of arbitrariness review under the APA, or else by interpretation of agencies' organic statutes.96 We argue

that judges have no warrant for so requiring. The Court has emphatically banned judges from imposing decision-

procedures on agencies as a matter of federal common law, and there is no source of positive law that might be

read to impose a global, judicially enforceable mandate of quantified CBA on agencies, even presumptively. (Later

we will take up the very different issue of quantified CBA mandates during executive review; such mandates by

their terms are not enforceable in court.)97

Quantified CBA is a particular decision-procedure-emphasis on procedure. Under Vermont Yankee98 and Perez,99

agencies have discretion about whether to adopt such procedures, and courts have no power to force them to do

so, beyond the procedures prescribed by the APA itself. Suppose that an agency is charged with deciding whether

to regulate mercury emissions from electric utilities,100 and suppose also that the court tells the agency that-even

after the agency has considered the relevant statutory factors- the agency must, if possible, quantify and monetize

the various factors. What the court has done is to prescribe how, and by what procedures, the agency is to exercise

its discretion. That is the very thing the Court has "[t]ime and again . . . reiterated" that lower courts are forbidden

to do.101

The only way around the Vermont Yankee problem is to locate the mandate for cost-benefit analysis either in

organic statutes or in section 706 itself, but neither approach succeeds. Where agency organic statutes are silent

or ambiguous, it is wildly implausible that Congress intends (or could be deemed fictionally to intend) a global

default rule requiring cost-benefit analysis. Across the broad landscape of federal regulatory statutes, Congress

sometimes mandates quantified CBA, sometimes refers vaguely to considerations of "cost," sometimes leaves

matters ambiguous, sometimes contents itself with Delphic silence, sometimes explicitly forbids consideration of

cost, and sometimes mandates other decision-procedures altogether, such as "feasibility" analysis.102 There is no

legal basis to elevate one of these approaches to global default-rule status, apart from a sectarian preference for

one approach or the other. The regulatory system writ large, like the Administrative Procedure Act, is a series of

"compromises" that allows "opposing social and political forces" to "come to rest."103 There is irreducible

reasonable disagreement about the best regulatory decision-procedure, including the very plausible view that there

is no single best decision-procedure, independent of context.

The same statutory landscape militates strongly against reading a (presumptive) requirement of quantified CBA

into the "arbitrary and capricious" language of section 706(2)(A). To do so implies that Congress itself acts

irrationally whenever it mandates feasibility analysis, or forbids cost considerations, as it sometimes does. Are the

proponents prepared to invalidate as irrational all statutes that mandate feasibility, or forbid considerations of

cost, perhaps under the Due Process Clause? Invalidate the Occupational Safety and Health Act, perhaps?104 (It is

no answer to say that the standard of rationality review is more deferential for Congress, although that is true.105

The view at issue is that not doing quantified CBA, when that is possible, is no more rational than using a Ouija

board-unconstitutional under any standard.) Whatever the answer to those questions, such provisions show that

many presumptively reasonable legislators, at various times, have not thought it irrational to mandate decision-

procedures other than quantified CBA. The proponents' reading of 706(2)(A) would thus produce an immediate and

severe incoherence in the federal regulatory system overall.

c. Current Law

So there is no basis for either a global default rule requiring quantified CBA whenever statutes are silent or

ambiguous, or else for reading an obligation to perform quantified CBA into the arbitrariness standard of 706(2)(A).

Fortunately, current law emphatically rejects both ideas in any event. The Supreme Court has consistently held

that quantified CBA is discretionary for agencies.106 There are a few exceptions that are memorable precisely

because they diverge from the normal course of judicial practice. But a brace of recent cases has clarified the

PDF GENERATED BY SEARCH.PROQUEST.COM Page 12 of 39

terrain.

In Entergy Corp. v. Riverkeeper, Inc.,107 the Court upheld EPA's use of cost-benefit analysis in the face of statutory

silence with respect to standards in section 316(b) of the Clean Water Act. The Second Circuit had held that the

EPA was not permitted to use cost-benefit analysis in determining the content of regulations under section

316(b).108 At issue in the case were EPA regulations concerning the technology required for operators of large

power plants that utilize "cooling water intake structures," drawing water in from water sources to cool the plant, in

the process killing aquatic life. The Clean Water Act requires that

[a]ny standard . . . applicable to a point source shall require that the location, design, construction, and capacity of

cooling water intake structures reflect the best technology available for minimizing adverse environmental

impact.109

So-called closed-cycle cooling systems recirculate the water used to cool the facility and therefore extract less

water from the local water source and generate less risk of harm to aquatic life. The EPA's rule required new

sources to use technology approximating the performance of closed-cycle cooling system. For certain classes of

existing sources, however, the EPA required reductions in the harm to aquatic organisms, but nowhere near the

reduction that would be accomplished if the EPA mandated performance at the closedcycle cooling system

level.110 While the closed-cycle cooling system performance would reduce impingement and entrainment

mortality by up to 98 percent, the costs of compliance for these existing sources would be nine times greater ($3.5

billion) than reducing aquatic damage by 80-95 percent using the alternative performance standards.111 Put

informally, the EPA concluded that the marginal costs of achieving the best possible reduction in aquatic harm

drastically outweighed the corresponding marginal benefit.

Writing for the majority, Justice Scalia explained that the EPA could permissibly read "best" to simply mean "most

advantageous."112 One possible interpretation of "best" is the technology that achieves the greatest reduction in

adverse environmental impact.113 Another, however, is the technology that "most efficiently produces some

good."114 The Court quickly dispensed with the notion that statutory silence necessarily precludes or mandates

cost-benefit analysis. On the contrary, discussing several cases in which statutory silence was read to preclude

cost-benefit analysis,115 or at least to not require it,116 the Court explained that "under Chevron, that an agency is

not required to do [cost-benefit analysis] does not mean that an agency is not permitted to do so."117 Even though

other standards in the CWA might preclude cost-benefit analysis in express terms, the EPA was free to adopt a

reasonable interpretation of the statutory standard above as allowing cost-benefit analysis. Far from announcing a

default rule in favor of cost-benefit analysis, the Court made clear that statutory ambiguity will generally be read to

give agencies discretion with respect to whether or not to utilize it.

Environmental Protection Agency v. EME Homer City Generation L.P.118 is similar. The case involved a challenge

to the EPA's interstate air pollution rules governing the conduct of upwind states that contribute significantly to air

pollution in downstream states. The Clean Air Act requires States to eliminate "amounts" of pollution that

"contribute significantly to nonattainment" in downwind states.119 But because multiple states may affect

conditions in a downstream state, the statute requires the EPA to apportion reductions. The D.C. Circuit had held

that the statute required EPA to allocate responsibility for reducing emissions proportionally to each State's

physical contribution.120 The EPA's method of determining cost-reduction obligations balanced the magnitude of

the state's contribution to the pollution of downstream air conditions and the costs associated with reducing

them.121 The challengers argued that the EPA was forbidden to consider costs, but, as in Entergy, the majority

again clarified that unless the statute clearly mandates otherwise, the EPA is free to consider costs. "The Agency

has chosen, sensibly in our view, to reduce the amount easier, i.e., less costly, to eradicate, and nothing in the text

of the Good Neighbor Provision precludes that choice."122 In short, in the face of statutory silence, the agency is

free to use cost-benefit analysis or not, as it sees fit.

In Utility Air Regulatory Group v. EPA123 ("UARG") the Court considered a challenge to the EPA's interpretation of

the Clean Air Act as it pertains to greenhouse gases (GHGs). In Massachusetts v. EPA,124 the Court had held that

greenhouse gas emissions were "pollutants" for at least some purposes of the Clean Air Act. After Massachusetts

PDF GENERATED BY SEARCH.PROQUEST.COM Page 13 of 39

v. EPA, the agency issued greenhouse gas emission standards for new motor vehicles.125 Stationary sources are

governed by two separate provisions of the Clean Air Act (as relevant here)-the Prevention of Significant

Deterioration (PSD) provisions126 and the Title V permitting program.127 Those provisions set numerical triggers

for what sort of entity is regulated (100 or 250 tons per year of a pollutant).128 Because greenhouse gases, unlike

other air pollutants, are emitted in vastly greater amounts, those numerical thresholds would have brought

thousands of new small entities under the rubric of the EPA permitting process.129 Accordingly, the EPA sought to

cover initially only those entities emitting more than 100,000 tons of greenhouse gases per year.130 In short, the

Court in Massachusetts v. EPA prodded the agency to adopt an ambitious reading of the statute, which then

threatened to render part of the existing regulatory framework unworkable.

Without delving too deeply into the technical details of the Clean Air Act, there were essentially two types of

sources at issue in the EPA rule challenged in UARG. The first group would not have been subject to the PSD or

Title V permitting process at all, but for their greenhouse gas emissions. The Court held that the EPA's

interpretation, which brought these thousands of sources into the Act's coverage, was unlawful.131 The second

group of sources were subject to the permitting process for the emission of other air pollutants anyway. For this

group of sources, there was a subsequent interpretive question about whether the EPA's decision to require Best

Available Control Technology (BACT) for greenhouse gases was a permissible interpretation of the statute. The

Court held that it was.132 The challengers essentially argued that BACT does not work for greenhouse gases.133

The Court's discretion-preserving language is striking: "applying BACT to greenhouse gases is not so disastrously

unworkable, and need not result in such a dramatic expansion of agency authority, as to convince us that EPA's

interpretation is unreasonable."134

Because UARG strikes down an EPA interpretation, it might at first glance be read as an aggressive form of judicial

review in tension with Entergy and EME Homer. In reality, however, it is precisely the opposite. EPA had advanced a

view that the agency was required to adopt the same definition of "air pollutant" throughout the Clean Air Act, and

therefore GHGs needed to be treated as criteria pollutants. As Justice Scalia explained, however, "the presumption

of consistent usage readily yields to context, and a statutory term-even one defined in the statute-may take on

distinct characters from association with distinct statutory objects calling for different implementation

strategies."135 Because the EPA's rigid interpretation would render the statutory scheme unworkable, it was not a

permissible interpretation of the statute. But where the interpretation would not render the scheme unworkable

(for "anyway" sources), EPA was free to adopt either interpretation. Indeed, this latter part of the opinion is striking

for just how far removed from mandatory cost-benefit analysis it is. The agency was not required to pick a rule that

was even close to cost-benefit justified so long as it was "not so disastrously unworkable."136

Finally, there is the recent decision in Michigan v. EPA,137 in which the Court invalidated an EPA rule under the

Clean Air Act relating to hazardous emissions from power plants. The relevant statutory text authorized regulation

only if "appropriate and necessary."138 As framed by the litigation, the question was whether the EPA could defer

consideration of costs at the stage of deciding whether to regulate, and later take costs into account at the stage

of deciding how much to regulate-how stringent regulation should be.139 In an opinion by Justice Scalia, the Court

held that under the statutory text, cost was a "relevant factor," and the EPA's decision to ignore cost at the first

decisional stage was unreasonable.140 The decision is most easily and naturally read as a Chevron Step Two

decision, on a straightforward statutory issue under the particular scheme of the relevant Clean Air Act

provisions.141

Proponents of quantified cost-benefit analysis point to seemingly broad language in the opinion,142 as when the

majority opined that "[o]ne would not say that it is even rational, never mind 'appropriate,' to impose billions of

dollars in economic costs in return for a few dollars in health or environmental benefits."143 On the broadest

possible reading, this could mean that it is arbitrary and capricious for agencies not to conduct quantified and

monetized cost-benefit analysis where possible. Yet this is an interpretation the Court took pains to disavow later

in the opinion. Justice Scalia went out of his way to emphasize that while rationality may require "paying attention

to the advantages and the disadvantages of agency decisions,"144 that is not the same as requiring quantification

PDF GENERATED BY SEARCH.PROQUEST.COM Page 14 of 39

of the advantages and disadvantages:

The Agency must consider cost-including, most importantly, cost of compliance-before deciding whether

regulation is appropriate and necessary. We need not and do not hold that the law unambiguously required the

Agency, when making this preliminary estimate, to conduct a formal cost-benefit analysis in which each advantage

and disadvantage is assigned a monetary value. It will be up to the Agency to decide (as always, within the limits

of reasonable interpretation) how to account for cost.145

Decisions about quantification, in other words, are a matter for reasonable agency discretion, contra Business

Roundtable. Michigan v. EPA is clearly alert to the distinction between the colloquial, informal sense of "costs and

benefits," on the one hand, and formalized quantified and monetized costbenefit analysis on the other. The

decision is principally an interpretive holding, about the meaning of the phrase "appropriate and necessary" in a

particular section of the Clean Air Act. But insofar as it addresses issues of rationality review in passing, it stands

only for the unobjectionable proposition that rationality requires consideration of both the "advantages and the

disadvantages of agency decisions."146 Absent specific statutory instruction, agencies may not conduct a one-

sided assessment of either costs or benefits without the other, any more than they may consider only one side's

factual evidence. The case stands for nothing broader than that unexceptional proposition.

d. "State Farm with Teeth"

Finally, there is an emergent idea in the commentary that tries to link a more intensive version of arbitrariness

review with the performance of costbenefit analysis by agencies. Professor Catherine Sharkey's proposed "State

Farm with Teeth" argues for a more intensive elevated standard for judicial review of some independent agency

decisions, particularly cost-benefit analysis not reviewed by OIRA.147 The core idea is that courts should review

agency decisions backed by high-quality cost-benefit analysis less intensively than they otherwise would. Judicial

practice along these lines would not expressly require cost-benefit analysis, but would do so indirectly.

To the extent that there is an implicit claim in Sharkey's proposal that courts should adopt a thin version of

rationality review for agency decisions that have been subject to OIRA review, we certainly agree. But to the extent

that she advocates a thicker version of review for agencies that have not engaged in rigorous cost-benefit analysis,

the idea is neither an accurate description of existing judicial practice, nor in our view a desirable shift in doctrine.

The proposal ignores the fundamental feature of agency decisionmaking under conditions of severe uncertainty.

To illustrate, agencies addressing health risks from pollution rarely know the exact shape of the doseresponse

curve at low levels of exposure.148 The challenge is to utilize other available information, for example effects of

exposure at higher levels or effects of exposure on animals, to make a reasoned inference about the likely effects

at low levels of exposure.149 So too agencies regulating financial markets will rarely be able to precisely state the

costs or benefits of a proposed rule because the actual effect will depend on a complex set of interdependent

decisions by market participants. The problem to be solved is a lack of certainty about those effects. To require an

agency to justify its decision using the exact information that is inevitably lacking is the very opposite of

rationality. It is akin to requiring that all requests to learn an unknown foreign language be made in that language,

or that all research be funded only if the results are already known.

But we have now moved to the question of agency decisionmaking under uncertainty; let us take up that topic

directly.

3. Connecting Facts and Choices: Uncertainty, Rationality, and Arbitrariness

In a frequently quoted passage, State Farm announced that an agency must "examine the relevant data and

articulate a satisfactory explanation for its action including a 'rational connection between the facts found and the

choice made.' "150 This has become a basic principle of rationality review: agencies must explain their choices, in

light of the facts. We will suggest, however, that there is much less to this requirement than meets the eye- both as

a matter of the theory of rational decisionmaking, and under current law. The obligation to explain choices, given

the facts, is far less demanding than lower courts sometimes assume-although the Court itself has usually

understood the problem and followed the correct approach.

The critical problem is that facts sometimes underdetermine agency choices. It is not necessarily the case, and

PDF GENERATED BY SEARCH.PROQUEST.COM Page 15 of 39

perhaps not even usually the case, that given some state of the world, the agency will always have (let alone be

able to give) reasons for choosing one policy over all competitors, or over any given alternative. Rather, it can be

the case, and may often be the case, that agencies will face a situation in which (1) the agency is obligated to

choose; (2) there exists more than one policy that can be justified, given the best evidence about the state of the

world; and yet, (3) there is no decisive reason to choose one policy over another.

Cases of this sort arise under genuine uncertainty, so-called Knightian uncertainty.151 Knightian uncertainty

arises when the decisionmaker has no respectable epistemic basis for attaching probabilities to possible

outcomes. This is not to say that subjective probabilities cannot be elicited and then attached, by brute force, but

those probabilities lack any credible epistemic warrant or foundation.

From the fact that it is always possible to elicit these subjective probabilities, we should not conclude that one

ought rationally to act upon them. One could certainly elicit from a political scientist the subjective probability that

he attaches to the prediction that Norway in the year 3000 will be a democracy rather than a dictatorship, but

would anyone even contemplate acting on the basis of this numerical magnitude?152

Furthermore, it has been shown empirically that subjective probabilities are highly sensitive to the method used to

elicit them, implying that they are artifactual.153

Absent probabilities, what is the rational decisionmaker to do? The main issue is whether the decisionmaker

should adopt a more or less optimistic or pessimistic approach. At the extreme of pessimism is "maximin," which

indicates the choice with the best worst-case outcome; at the extreme of optimism is "maximax," which indicates

the choice with the best bestcase outcome; and indeed any weighted average of these extremes is also

possible.154 Importantly, any of these approaches is equally rational, given the circumstances; there is no general

way to arbitrate among them, no further requirement of rationality that would knock out all but one approach.

Sometimes, of course, law will impose further constraints, even if the theory of rational decisionmaking does not.

Particular regulatory statutes might command a highly cautious or pessimistic ("precautionary") approach. But

there is no general requirement to that effect; the law rejects any general preference for maximin over maximax.

The Supreme Court has consistently overturned judges who attempt to impose on agencies a mandate to make

"conservative" or "worst case" assumptions under uncertainty.155 There is thus an element of irreducible

discretion in agency choice under genuine uncertainty. The facts underdetermine the choice, and as far as the law

and the theory of rational choice goes, the agency can simply choose how pessimistic or optimistic to be.

In an important set of cases, then, agencies face a dilemma: although the agency does best by choosing one of the

available options from within the feasible set, any choice of a particular option is necessarily arbitrary. The agency

thus faces a "rationally arbitrary decision"156: it is rational to act arbitrarily, in the sense of making a policy choice

that cannot be justified relative to other available choices, but can be justified by the need to make some choice or

other from within the feasible set. Rationality and arbitrariness are usually cast as strict antonyms, but this is a

confusion. Agencies acting under uncertainty may have perfectly good second-order reasons to make one choice

or another, within the feasible set, even if they lack any first-order reason to choose one option over another within

the feasible set. In cases of this sort, agencies act based on reasons, even if they have no first-order reasons for

their choices; they act arbitrarily, but have perfectly good reason to do so.

As a recent illustration, the Secretary of the Interior was required by law to decide whether to list the southwestern

flat-tailed horned lizard as a threatened species or not, yet there was no reliable data on the number of extant

lizards;157 "the administrative record did not support a finding that the lizard population was viable or

nonviable."158 In such circumstances, the agency has excellent second-order reasons-here, compliance with a

legal mandate-to make a decision one way or another, even though no decision can be fully justified.

Courts should recognize the existence of this sort of dilemma. The hallmark of such cases is mirror-image

reversibility: the agency's choice of A over B is arbitrary, in the sense that the agency can give no valid first-order

reason for the choice, but it is equally true that the agency can give no valid reason for the opposite choice either.

The agency is constrained to choose, but any choice it makes would fail ordinary arbitrariness review. Recognizing

this dilemma, the court should defer to the agency's choice.

PDF GENERATED BY SEARCH.PROQUEST.COM Page 16 of 39

Any other judicial approach will create deadweight losses for the system. If the court itself picks arbitrarily,

substituting its own decision for the agency's, there is by hypothesis no added benefit whatsoever, but there is the

extra cost of the judicial proceeding itself, including not only the out-ofpocket costs but also the delay in reaching

some choice or other. Furthermore, as we will discuss shortly under the rubric of tacit expertise, the agency's

choice may actually be better than the court's in ways and on grounds that the agency cannot explain to the court;

the agency may simply have better instincts, even in situations of severe uncertainty. If, on the other hand, the

court demands first-order reasons for the agency's choice, then any of the agency's possible choices can be

overturned as arbitrary. On that approach, one of two pathological results will occur. Either the agency will be

trapped in an indefinite cycle of reversal, as the court overturns every new agency attempt to make a policy choice,

or-more likely-agencies will have powerful incentives to engage in a charade, in which they offer the reviewing

court bogus first-order reasons to prefer one alternative over the other. In such situations, it is pathological for

courts to relentlessly demand that the agency supply first-order reasons that in the nature of things cannot be

given. At the frontier of uncertainty, rationality simply runs out.

Two doctrinal conclusions follow, one involving Section 706(2)(A) of the APA, the other involving State Farm and

the rational connection test. The first conclusion is that decisions may be arbitrary in a theoretical sense without

being arbitrary in a legal sense. In cases involving mirror-image reversibility, agency decisionmaking may be

rationally arbitrary but not legally arbitrary.

The second conclusion is that in cases of this sort, agencies may validly select policies without a direct

connection between the facts found and the choices made. Because the facts and the state of the world

underdetermine choices, agencies cannot reasonably be asked to show, based on the facts, that the choice they

make is superior to the alternatives. This is not to say that there is no rational connection between facts and

choices. It is to say that the nature of the rational connection is different than in standard cases. The rational

connection lies at the second order, not the first; it arises because the agency has good reason to decide, even if it

lacks good reason for the decision. In this sense, the State Farm test of rational connection is more capacious,

more forgiving, and less demanding than is conventionally understood.

So far we have said nothing at all about the empirical incidence of such cases. Our claim is analytic, and will hold,

or not, regardless of the nature of the problems that actual agencies actually face. Still, if only to underscore the

importance of getting the right conceptions of rationality and rationality review, it is worth recording our sense of

the matter.

On one view, which is certainly possible and which cannot be refuted on a priori grounds, cases of genuine

uncertainty are rare in the field of agency decisionmaking; agencies mostly encounter problems of risk, to which

quantifiable probabilities can be attached, and to which quantified cost-benefit can be applied. Our view is

different. Agencies frequently encounter novel problems at the frontiers of scientific and technical knowledge,

such that even expert probability assessments are unreliable and real uncertainty is pervasive. The ordinary

problems of risk assessment and regulation, involving calculable probabilities and known risks, have a diminishing

market share over time, as agencies handle easy problems and move on to more difficult ones. When agencies

face regulatory choices premised on guesses about the effects of climate change,159 or the chances of a novel

type of domestic terror attack,160 or the future of a species whose number is simply unknown,161 or the effects of

a novel regulatory constraint on shareholder voting,162 they face choices in which any probability assessments

lack respectable epistemic foundations-they face genuine uncertainty. A sensible theory of rationality review has

to take account of the pervasive presence of uncertainty in the administrative state.

4. Comparative Policy Evaluation (Over Time)

Uncertainty has another critical dimension, involving the rationality of information-gathering by agencies. Rational

agencies will invest resources in acquiring information, which may resolve Knightian uncertainty, transforming it

into risk or even certainty.163 Yet "in some cases the value of further investments in information gathering will

itself be genuinely uncertain."164 If so, the problem of rational arbitrariness will replicate itself at this higher level

as well.165 In a genuinely uncertain choice environment, the decisionmaker must stop the search for the best

PDF GENERATED BY SEARCH.PROQUEST.COM Page 17 of 39

policy sooner or later, somewhere or other, or else fall into an infinite regress-deciding whether to acquire

information about the costs of acquiring further information, and so on.166 Policymaking under uncertainty is

like going into a big forest to pick mushrooms. One may explore the possibilities in a certain limited region, but at

some point one must stop the explorations and start picking because further explorations as to the possibility of

finding more and better mushrooms by walking a little bit further would defeat the purpose of the hike. One must

decide to stop the explorations on an intuitive basis, i.e. without actually investigating whether further exploration

would have yielded better results.167

In administrative law, the problem of information acquisition arises most critically with respect to arbitrariness

review, particularly the scope of agencies' obligation to consider and evaluate policy alternatives. We have seen

that although both State Farm and successor cases explicitly repudiate the idea that agencies must consider all

feasible policy alternatives,168 many judges act as though there is such an obligation, often without quite saying

so. The underlying impulse here is the tempting thought that comparative policy evaluation is a necessary element

of rational decisionmaking, for agencies or indeed for any institution or actor. Surely, the intuition runs, rationality

requires choosing the best option, relative to the chooser's preferences, within the feasible set of choices. Suppose

there is an agency charged with reducing air pollution, at acceptable cost; and suppose the agency were to say

that "policy P is cheaper than policy Q, and P yields more reduction in air pollution. But Q isn't bad at all; we think

it's good enough. We choose Q." Isn't that irrational?

Judges who reason this way have an entirely legitimate intuition, but they fail to realize that the intuition does not

always hold, and that the conditions under which it fails to hold are especially likely to arise in the administrative

setting. The underlying issue is the validity of "satisficing"-of picking something on the ground that it is "good

enough."169 Satisficing is intrinsically noncomparative. The satisficer picks a feasible option whose quality

exceeds some predefined aspiration level, regardless of whether there might be an even better option somewhere

in the feasible set.170 In real life, people constantly satisfice; indeed, people who relentlessly seek the best

possible option have a mad air about them. The ubiquity of satisficing should alert judges that comparative policy

evaluation is an approach that makes sense only under particular conditions.

What are those conditions, and what are the conditions under which satisficing is sensible? Satisficing becomes

sensible in the presence of substantial costs of information and search-in a word, under uncertainty.171 In the

earlier example, the agency's choice of Q over P was irrational only because the context of choice was entirely

static and transparent; the options were known, as were their full costs and benefits. In static contexts, absent

uncertainty, satisficing is indeed irrational and comparative policy evaluation is indispensable.172 In more

dynamic contexts, however, satisficing comes into its own. The satisficing decisionmaker applies a stopping rule

that constrains the open-ended search for the very best policy, in favor of one that meets or exceeds the aspiration

level.173

The crucial twist here is that, while satisficing is a rational strategy for agents with limited time and information

who are forced to make choices under uncertainty, the selection of an aspiration level is itself inescapably

arbitrary. Nothing in the idea of satisficing, or in the choice situation itself, tells the rationally satisficing agent

where to locate the aspiration level, higher or lower. In that sense, satisficing is a strategy of rational a-rationality,

of rationally arbitrary decisionmaking. It is an approach that seemingly lacks any justification in first-order reason-

an observer may well complain "why did you stop there, not somewhere else?"-but it is justified as a strategy, over

an array of problems, by higher-order considerations.

All this is important because uncertainty pervades regulation and other forms of agency policymaking. In the

administrative setting, choices are rarely fully-specified, static and transparent. The chronic condition of agency

policymaking is the search for sensible policies under uncertainty. At some point, the agency will have to suspend

the search and choose something good enough, even if it is abstractly possible that there exists a better policy

that is technically feasible. Administrative law doctrine, at its best, recognizes exactly this point by underscoring

that agencies' obligation to consider policy alternatives is limited to reasonable alternatives.174 The adjective

represents an implicit recognition that consideration of technically feasible alternatives is often a game not worth

PDF GENERATED BY SEARCH.PROQUEST.COM Page 18 of 39

the candle. Even in State Farm, the Court was careful to specifically deny that agencies have any obligation to

"consider all policy alternatives in reaching decision."175

That denial has sometimes been forgotten by lower federal courts. At its worst, intrusive judicial review threatens

to create an infinite regress, in which agencies have to be able to give reasons for suspending the search for

optimal policies, reasons that require the very information whose absence is the reason for stopping in the first

place. "The reason agencies do not explore all arguments or consider all alternatives is one of practical limits of

time and resources. Yet, to have to explain all this to a reviewing court risks imposing much of the very burden that

not considering alternatives aims to escape."176

As in other settings, however, the Supreme Court often displays a better understanding of uncertainty and its

significance for administrative law than do the lower courts.177 Thus Baltimore Gas allowed the Nuclear

Regulatory Commission to adopt a maximally optimistic assumption about the environmental effects of spent

nuclear fuel in the remote future, an assumption of "zero-release."178 In our terms, Baltimore Gas recognized that

under uncertainty maximax is just as valid as maximin.179 More importantly still, FCC v. Fox180 squarely held that

comparative policy evaluation is not a general requirement of rational agency decisionmaking. Agencies have no

general legal obligation, as far as the APA is concerned, to show that the chosen policy is the best among the

feasible alternatives, relative to the agency's stated goals. As the Court put it,

[T]he agency must show that there are good reasons for the new policy. But it need not demonstrate to a court's

satisfaction that the reasons for the new policy are better than the reasons for the old one; it suffices that the new

policy is permissible under the statute, that there are good reasons for it, and that the agency believes it to be

better . . . .181

The obligation to show that "there are good reasons for the new policy," coupled with the lack of any obligation to

show that chosen policy is better than the alternatives, is in effect a satisficing approach, rather than a

comparative one.

A wrinkle in Fox is that the issue of comparative policy evaluation in the case involved a change in agency position

over time. The case involved a change of agency policy; the principal dissent, by Justice Breyer, urged that

agencies should have to show the comparative superiority of the new policy-a somewhat puzzling stance, given

Breyer's usual sensitivity to the impossible burdens that a requirement of full comparative evaluation would

impose.182 The Court rejected Breyer's view as too demanding. The Court observed-citing State Farm, quite

correctly-that rationality review requires agencies to take into account data that can "readily be obtained," but does

not require "obtaining the unobtainable."183 To be sure, the Court added critical qualifiers that close some of the

distance between majority and dissent. The Court, for example, acknowledged that agencies should explain

changes in factual assertions, and should take into account knowable and known costs of transition to the new

policy ("reliance interests").184 But there remains an irreducible difference between majority and dissent: the Court

is very clear that comparative policy evaluation, in and of itself, is not a requirement of rationality review.185

The Fox principle, denying any agency obligation to engage in comparative policy evaluation, fits perfectly with the

principle that agencies may proceed one step at a time, enjoying "broad discretion" to parcel out policy questions

across different proceedings, present and future.186 The two principles actually entail one another. Because

agencies need not consider all relevant alternatives now, they cannot have any obligation to show that the

currently chosen action is superior to the relevant alternatives, which may be allocated to a separate proceeding

entirely, or simply put off the table for the time being. The point common to both principles is that agencies need

only have some adequate reason for what they do now; they need not rank possible actions and show that the one

chosen is best.

Whether the issue is comparative evaluation of feasible policy choices at a given time, or over time, the point is the

same. Under uncertainty, in the presence of costs of information and search, rationality does not require agencies

to show that they have chosen the best of the technically feasible alternatives. Thanks to Fox, the prevailing law of

rationality review does not require that either.

5. Accuracy vs. Other Values: Tradeoffs and Decision Rules

PDF GENERATED BY SEARCH.PROQUEST.COM Page 19 of 39

Lurking in the background of thick rationality review is an assumption that arbitrariness review requires the agency

to make the best decision, in the narrow sense, identifying the optimal policy response to the complete set of facts

that are known or could be discovered. We have already shown that this idea is inconsistent with existing Supreme

Court doctrine, which neither requires optimal cost-benefit balancing nor requires that the agency identify and

adopt the best of all feasible policy alternatives.187 There are also a range of other conditions under which, for

good second-order reasons, rationality review does not require that the agency pick the best expected policy, in the

first-order sense.

a. Mean-Variance Tradeoffs

A standard problem for agencies (and all decisionmakers) involves mean-variance tradeoffs. When making

predictions about the effects of different policies, agencies must predict the most likely or average effect of a

policy. The agency's "best guess," for example, might be that changing the National Ambient Air Quality Standards

(NAAQS) for ozone will save 10,000 lives. But that estimate-the average or mean-has a variance as well. The better

the information or data the agency has, the lower the variance. The more speculative or uncertain the information,

the higher the variance. Just to illustrate, a good estimate might entail a variance of plus or minus 1,000 lives; a

lower quality estimate might entail a variance of plus or minus 8,000 lives.188

Assuming the agency has already gathered all cost-justified information, it is entirely rational to pick the policy

with the lower expected number of lives saved if that policy option also produces a smaller variance. Take an

extreme example for the sake of crisply illustrating the point. If policy A is expected to save 10,000 lives, plus or

minus 9,000, and policy B is expected to save 9,999 lives, plus or minus 1, rationality does not require policy A, and

it might very well require the opposite. To give a more banal example, suppose that one's ideal temperature is 72

degrees. When choosing between two potential vacation locales, it is entirely rational to choose spot A with an

average temperature of 70 degrees plus or minus 2 degrees, instead of locale B which has an average temperature

of 72 degrees plus or minus 20 degrees. There are many contexts in which variance is more important than the

mean, and at a minimum, it is well within the bounds of rationality-on any standard conception-to prefer locale A as

a vacation destination to locale B. Agencies face analogous decisions constantly. A strong form of rationality

review that requires agencies to pick policy alternative with the highest expected return would be seriously

misguided. Mean-variance tradeoffs exist in almost any agency decisionmaking setting and agencies should not

be required to ignore them by a confused formulation of rationality review by courts.

b. Speed vs. Accuracy

Increasingly, agencies are asked to formulate new rules under significant time pressure. The timelines that Dodd-

Frank established for the promulgation of hundreds of new rules regulating the financial industry were

astonishingly short, and as a consequence most of the deadlines were missed by the agencies.189 The Food and

Drug Administration was successfully sued because it failed to meet the aggressive time-frame established by the

Food Safety Modernization Act.190 Statutory deadlines respond, in part, to longstanding criticism in the

commentary that the pace of agency rulemakings was too slow.191

Whether or not agencies, in fact, tend to move too slowly, it is clearly true that that agencies are often under

pressure to act more quickly. In the case of true emergencies, the "good cause" exception of the APA will allow

agencies to avoid notice-and-comment requirements at least for a brief time period.192 Yet this is only relevant to

a particular subset of agency action, involving legislative rulemaking; there are many other cases in which speed is

of the essence. Dollars or lives may be lost if agencies move slowly. In such cases, agencies face an inevitable

tradeoff between speed and accuracy.

The speed-accuracy tradeoff is an important special case of the costs of information-gathering. With more careful

study, consultation, the design of new experiments, the construction of new models, and so on, agencies could

more accurately identify the best policy. Yet rationality will sometimes dictate acting quickly, trading accuracy for

speed, because of the costs of waiting. Indeed, a rule that always favors more accurate decisionmaking and

ignores temporal problems would arguably be irrational-not to mention that it would result in virtually no agency

action ever. As should be immediately clear, this is a problem for thicker forms of arbitrariness review, which

PDF GENERATED BY SEARCH.PROQUEST.COM Page 20 of 39

require a strongly rational connection between the facts known and the choice made.193 It is often rational, indeed

optimal, to not spend the time gathering information so that a clear rational connection exists between particular

facts and the particular choice made, because that would require sacrificing the benefits of expedition. This sort of

tradeoff does not always exist, but it sometimes does exist, and arbitrariness review should be flexible enough to

accommodate.

c. Asymmetric Error Costs

Administrative law doctrine on arbitrariness review tends to obfuscate the problem of asymmetric error costs.

Criminal law has long been focused on these sorts of costs. The Blackstone principle-that it is better to let ten

guilty persons go free than to convict one innocent person-is a potential justification for many defendant-favoring

rules of criminal procedure.194 This principle is an outgrowth of the problem of asymmetric error costs: it is much

worse to err by putting an innocent person in prison than to err by putting a guilty person on the street, or so the

argument goes. Irrespective of whether one agrees with this intuition,195 problems of asymmetric error costs

abound in the law. Decisions about whether to list endangered species have this flavor. Failing to list a species

that should have been protected will result in extinction; listing a species that didn't need to be protected will result

in some financial costs but can be corrected at some later point. Failing to regulate a new pollutant that should

have been regulated will result in illness and deaths. Regulating a new pollutant unnecessarily will result in

financial costs.

The costs of making mistakes in administrative law are often asymmetric, and it is perfectly rational for agencies

to take asymmetric error costs into account. Even if the best available evidence suggests the best policy

alternative is X, it is rational for an agency to select policy Y because the costs of error are not symmetric. This is

one common justification for the "precautionary principle."196 Even if the agency has failed to pick the best policy

in a first-order sense, it has good second-order reasons for not doing so. The agency's decision rule is rational,

even if its narrow decision is not. It is an open question whether (lower) courts would accept this sort of

justification, but they should.

6. Tacit Expertise

Finally, the literature on rationality review has entirely ignored a crucial distinction between having reasons and

giving reasons.197 Here two different strands of thought about rationality review come to the fore. In one strand,

the desirable thing, the thing rationality review aims to produce, is to ensure that agencies have good reason for

what they do. On this conception, if reviewing courts (or other reviewing institutions) could somehow be assured

that agencies were in fact acting based on reasons, then it wouldn't matter if those reasons were never actually

communicated to courts. The communication itself has a strictly derivative and incidental value; it is a strictly

evidentiary mechanism, one that helps courts or other reviewers to flush out illicit agency motivations by

comparing the agency's actions to its stated rationales. But in principle the communication of reasons by agencies

could be dispensed with if the judges were otherwise assured that agencies were acting on the basis of valid

reasons.

On a very different conception, the act of public communication of agency reasons is itself valuable, independent

of its evidentiary function. The public communication may enable nonjudicial oversight of agencies, promote

democratic transparency, and give agencies ex ante incentives to formulate rational policies, in light of the

looming prospect that they will be obliged to explain the policies to third parties. On this conception, it is important

not merely that agencies have reasons, but that they give them.

These two conceptions will diverge entirely if there exists a class of cases in which agencies have reasons they

cannot give, explain or convey, at acceptable cost. Unfortunately, there is every reason to think that such cases not

only exist, but are common. Tacit expertise is a widespread phenomenon. In many fields, experts have tacit

knowledge that they cannot communicate, at least at acceptable cost, to generalist observers or other nonexperts.

Tacit knowledge is how-to local knowledge. In Hayek's terms, it is based on "particulars"-"circumstances of time

and place" that "by [their] nature cannot enter into statistics and therefore cannot be conveyed to any central

authority in statistical form."198 Because such knowledge is ubiquitous, there is always a question whether those

PDF GENERATED BY SEARCH.PROQUEST.COM Page 21 of 39

who can do something are able to explain how they do it. Those who can do often can't teach.

Although some tacit expertise may simply be incommunicable in principle (riding a bicycle is the stock example:

one can show how to ride, but not explain how to ride in the abstract), probably the more common case involves

tacit expertise that can be explained with sufficient time and effort. The issue is the cost of doing so. If the cost is

substantial, there will be cases in which forcing an explanation is simply not worth it. In such cases, that is,

decisionmakers will have legitimate tacit reasons for their choices, reasons that cannot be communicated to

reviewers at acceptable cost.

a. Tacit Expertise, Agencies, and Courts

In the administrative state, agencies have trouble explaining to generalist reviewers critical elements of regulatory

problems that line experts, acting in the field and saturated with local, field-specific knowledge, understand on an

essentially intuitive basis.199 Presumably for just this reason, in the intraexecutive setting, a critical OMB

guidance document on cost-benefit analysis observes that "professional judgment" is essential to high-quality

regulatory analysis, especially when nonquantifiable costs and benefits are at issue.200 Courts should certainly be

no more skeptical of expert professional judgment than is OMB. Judicial review should take into account the

existence of cases in which agencies possess tacit expertise and thus have reasons it is costly for them to give.

It goes without saying that the opposite cases also exist, in which agencies have no sensible reason for their

choices, or are making decisions based on biased information, or political motivations in the pejorative sense. In

such cases, the requirement to not only have a (public-regarding) reason, but to give such a reason, may prevent

arbitrary agency behavior. Our point is simply that these two types of cases are both real; both must be taken into

account by the institutional calculus. A review procedure that is oblivious to the phenomenon of tacit expertise will

code some set of cases as instances of arbitrary agency action, in the sense of unreasoned agency action, when in

reality they are cases of reasoned action that cannot be explained at acceptable cost. Again, it is a separate

question whether the best legal conception of rational agency decisionmaking should or should not code such

cases as "arbitrary." But without appreciation of the issue of tacit expertise, one cannot even recognize the critical

questions.

In the real world, a review procedure that is oblivious to tacit expertise will induce coping tactics and tactics of

self-defense by agencies confronted with generalist reviewers.201 Unable to convey the real, albeit tacit, grounds

for decision to the reviewers at acceptable cost, agencies will tend to substitute articulable reasons for their real

reasons. The articulable reasons will fit the agency's behavior less well than the true, tacit grounds of decision; the

result will be a kind of distortion in the process of agency reason-giving, a distortion arising from the limitations of

the reviewers. This is law's answer to the Heisenberg effect: the act of observing an agency's reasoning, through

hard look review, changes the nature of the agency's reasoning itself.

The distortion induced by the effect is a cost, insofar as the goal of the review system is to ensure that agencies

act based on expertise. If the system places some positive independent value on publicly communicable reasons,

then that distortion may or may not be a price worth paying; but one should at least be clear that a price is in fact

being paid. All we aim to establish here is that a system of review that is oblivious to the phenomenon of tacit

expertise will produce false positives. It will falsely code some agency decisions as cases of genuinely arbitrary

decisionmaking, when in fact the agency is acting on the basis of genuine expertise. To call such decisions

unreasoned is merely to underscore the limitations of the reviewers, not the reviewed.

b. The Pretext Problem and Some Rules of Thumb

Let us make the preceding intuitions somewhat more systematic. We have underscored the existence of cases in

which judicial recognition of tacit expertise would allow agencies to act on the basis of reasons they have but do

not give. This is, of course, only one side of the ledger. The other side is the problem of pretext. Judicial recognition

of tacit expertise implies the possibility that agencies may claim to have tacit reasons that do not actually exist.

The Type I error is judicial failure to recognize tacit expertise when it does exist, resulting in erroneous invalidation

of agency action; the Type II error is to deferring to nonexistent tacit expertise, resulting in erroneous validation of

agency action.

PDF GENERATED BY SEARCH.PROQUEST.COM Page 22 of 39

The overall picture may be sketched this way. An agency might have good or bad reasons (somehow defined) for a

decision and the giving of reasons to judges might generate good or bad effects, depending on the case. Consider

four possibilities:

(1) The agency has bad reasons (however defined) for its decision and requiring the agency to offer a rational

account of its decisionmaking catches the bad reasons, allowing the court to strike down the agency action as

arbitrary.

(2) The agency has bad reasons but can articulate a plausible pretext, so arbitrariness review fails to stop the

action, notwithstanding the reason-giving requirement.

(3) The agency has good reasons and honestly articulates them. Here, the requirement of reason-giving does not

fix a problem because there is no problem to fix, but it nonetheless imposes some costs perhaps modest, perhaps

extensive, because tacit expertise has to be represented to generalist judges.

(4) The agency has good reasons for its decision, but those reasons are grounded in tacit expertise, and are too

costly to communicate as such. A requirement of reason-giving makes it more likely a court will wrongly strike

down the action as arbitrary because the agency must articulate a false reason for a policy or imperfectly convey

the real reason for a policy.

Thick rationality review, requiring agencies not only to have reasons but to explain them to the satisfaction of

generalist judges, does well in case (1), but does poorly in cases (2), (3) and (4). For the reasons noted above, we

suggest that case (4) is plausibly quite common. After all, the motivating assumption for most of administrative

law is information asymmetry.202 If Congress or courts could easily learn all the agency knows, the information

asymmetry problem would recede, and agencies wouldn't really be necessary at all. If it is implausible that tacit

expertise can always or even usually be communicated to a reviewing court at reasonable cost, then thick

rationality review is likely to make matters worse, not better.

However, we need not establish, and do not claim to know-which review procedure is best overall. Our minimum

point for present purposes is merely that both the Type I error and the Type II error exist, and both are costly.

Administrative law's approach to judicial review should take both errors into account in an optimizing calculus,

minimizing the total costs of the two types. The challenge would be for law to develop principles, standards, or

rules of thumb that would allow courts to sort between cases where agencies are more or less likely to possess

genuine tacit expertise. How exactly this should work is a topic for another paper; tacit expertise is an idea that

deserves more extended treatment than we can give it here. We will limit ourselves to a brief preliminary sketch of

some rules of thumb for reviewing courts.

To be clear, these rules of thumb are just indirect proxies, and will thus be necessarily imperfect. They are proxies

because courts cannot, by definition, directly observe the state of agencies' tacit expertise to decide whether a

particular claim of tacit expertise is plausible. If such direct observation were possible, the whole problem would

disappear. All courts will be able to develop are imperfect sorting principles that help them to decide whether the

Type I or Type II error is more likely in the given case at hand.

Statistical versus experiential questions. The main proxy must always be the nature of the issue at hand. Recall

Hayek's distinction between local knowledge of "circumstances of time and place," on the one hand, and general

knowledge that can be expressed in statistical form for transmission to a central authority, on the other.203 Some

agency decisions turn on statistical facts-generalized legislative facts, subject to measurement and

methodological contestation. Other agency decisions turn on localized knowledge acquired through experience

with a particular industry or with particular actors within the industry, whether individuals or firms. At a certain

point, the United States stopped conducting live tests of new forms of nuclear weaponry, and turned to statistical

modeling of the likely effects of (hypothetical) detonations.204 Whatever its larger merits, the downside of this

approach was that only veteran engineers had personal experience with test detonations, and only they could

answer critical questions about how new weapons were likely to work in practice, as opposed to within

equations.205 So too, problems of industrial hygiene involve experience-based best practices, on which abstract

economic expertise gets little purchase.206 Tacit expertise is more or less likely to arise in different contexts, and

PDF GENERATED BY SEARCH.PROQUEST.COM Page 23 of 39

courts should be sensitive, above all, to the type of issue before them.

Character and traits. A special type of local knowledge-but a ubiquitous type-is knowledge about the character and

traits of individuals (and perhaps of firms, although there is a separate and interesting question about whether

firms have a character). Agencies must often make judgments based at least in part on the character of particular

people-CEOs, for example- with whom they interact. Over time, officials in agencies will come to understand, in a

quasi-intuitive way, who is to be trusted and who is not. As an example of this thick practical knowledge, consider

agency staff who possess the savvy, born of experience, to discern which automobile executives are bluffing about

the capabilities of their firms and about the costs to those firms of various regulatory outcomes.207 Such

knowledge will often be costly or downright impossible to document, in any convincing way, for a third party, such

as a court exercising judicial review. But it is a real form of knowledge nonetheless.

Agency reputation. Finally, the pretext problem is self-limiting, because agencies that constantly base their

decisions on (putatively) nontransmissible tacit expertise will encounter increasing skepticism from reviewing

courts over time. Notwithstanding the theoretical uniformity of standards of review, the D.C. Circuit sees a steady

stream of decisions from a recurrent set of the same agencies.208 Some agencies have credibility established

over many years; others suffer from the opposite problem. Just as certain advocates get more credibility and

perhaps the benefit of the doubt during oral argument, some agencies do as well. When this occurs, what appears

to be the a robust form of arbitrariness review is rather a form of escalating review, in which the intensity of

judicial investigation grows over time as an agency repeatedly declines to articulate justifications for its decisions,

beyond an appeal to tacit expertise.

We do not claim to have done justice to the formidably complex, yet indispensable, idea of tacit expertise. All we

have attempted to do here is to sketch some basic issues and possible lines of analysis, in order to open up new

questions for another time. For purposes of rationality review, all we need establish is our minimum point: courts

should take into account not only the risk of pretext, but the countervailing risk that agencies have tacit,

experiential reasons for action that they cannot explain to generalist judges (at least not at acceptable cost).

B. Agency Rationality: A Positive Formulation

So far we have been relentlessly negative. To some degree, negativity is inherent in the nature of our project,

indeed the very point of our project; our principal claim is that there is and should be less to rationality review than

some judges and commentators seem to think. Nonetheless, it is incumbent on us to articulate a positive

formulation of agency rationality and rationality review, in order to offer an alternative to the hard look approach

we reject.

In our positive formulation, the best interpretation of section 706(2)(A), and of rationality review, is simple indeed:

agencies must act based on reasons. In this simple conception, the aim of section 706(2)(A) is to exclude agency

action that rests on no reasons whatsoever, at any order of analysis- the core meaning of "arbitrary and

capricious." The key difference between our conception and the hard look conception is that the former, unlike the

latter, takes account of nonideal constraints on agency decisionmaking. It recognizes that limits of time,

information, and resources may give agencies good second-order reasons to act inaccurately, nonrationally, or

arbitrarily, in a first-order sense. In a particularly naive version of rationality review-a version that nonetheless

appears with some frequency in judicial opinions, usually in an implicit form-agencies must have fully specified

first-order reasons for their choices, reasons that justify their choices relative to all competitors, in light of the

statutory policy goals.209

Our conception opens up space for agencies to act based on reasons at a second-or-higher order. Such reasons

may, for example, justify a policy that seems acceptable, but might well be worse than other possible policies in

the feasible set (for satisficing); justify acting when taking some action or other is necessary or desirable, even

when no particular action is sufficiently justified (a rationally arbitrary decision); justify a policy under a decision

rule that can predictably be expected to misfire, producing arbitrary results, in some set of cases (the mean-

variance tradeoff and the speed-accuracy tradeoff). In all these cases, agencies rightly depart from the simplistic

benchmark under which rationality requires choosing the best option within the (known) feasible set. By parallel,

PDF GENERATED BY SEARCH.PROQUEST.COM Page 24 of 39

arbitrariness review must also depart from the simplistic idea that courts should require agencies to explain, in a

firstorder sense, why their chosen policies represent the best choice from the (known) feasible set.

Reasonableness and rationality. We have described this approach as a thin theory of agency rationality, because

agencies who act in the ways we have described are emphatically acting rationally at the second-or-higher order.

The nonrationality or arationality of their behavior, at the first order, is in the service of rational strategies for

coping with environments in which ordinary first-order rationality runs out or misfires. It is worth mentioning,

however, a different description of the problem that some find congenial, and that works equally well for our

purposes.

On this alternative description, we might distinguish reasonable agency action from fully rational agency action.

When agencies bump up against the limitations of rational choice, as when they (arbitrarily) pick an aspiration

level in order to satisfice under uncertainty, they are acting reasonably, even if not rationally. When the canons of

rational choice prove indeterminate or ambiguous, and fail to prescribe a unique choice under conditions of

uncertainty, the limits of rationality are reached. It does not follow, however, that chaotic or capricious

decisionmaking is the only alternative. Rather it is possible to decide reasonably, even when rationality has

exhausted its force. For many large decisions at the individual level-where to go to college, what profession to

pursue, whom to marry-rational choice is impotent or inapposite, yet it is still possible to approach the decision

more or less reasonably. Many of the decisions that agencies face have exactly this quality: the stakes are high,

the consequences of the alternatives are shrouded in uncertainty, and the decision is either a one-time event, or at

least will not be frequently repeated, so that no strong process of learning through trial and error is possible.

In this framework, our thesis holds that for purposes of interpreting section 706(2)(A), "arbitrary and capricious"

action is unreasoned agency action. Just as an individual may have excellent reasons to make a decision that

cannot be fully justified in rational terms, so too an agency may have excellent reasons to adopt a decision-

procedure (like satisficing) that is justifiable by reasons, but yields ultimately nonrational choices in particular

cases. Judges who appreciate the limits of rationality, and the dilemmas that face reasonable agencies who must

act subject to those limits, should interpret the APA's rejection of arbitrary decisionmaking in ways that take

account of these concerns.

C. Nonjudicial Review

So far we have mentioned two different types of reviewers: generalist judges, on the one hand, and technocrats in

the Office of Information and Regulatory Affairs, on the other. Our focus is on judges, and on judicial review for

rationality under section 706(2)(A), rather than OIRA review under relevant statutes and executive orders. For

completeness, however, it is worth asking which, if any, of our claims apply only to judicial review, and which apply

in the intraexecutive setting as well.

Some of our points are not sensitive to the identity of the reviewer. We suggest that the ideal of rational

decisionmaking by agencies, rightly understood, is more capacious and forgiving than is often assumed. We

counsel tolerance of noncomparative policy evaluation and satisficing under uncertainty; of rationally arbitrary

decisionmaking under uncertainty; of tradeoffs that compromise simple accuracy, such as mean-variance

tradeoffs and speed-accuracy tradeoffs; tolerance for agency discretion to decide which factors are relevant to the

goals of statutory policy; and tolerance for agencies who possess tacit expertise acquired through experience, and

who thus have reasons they cannot necessarily explain.

The problem of tacit expertise is worth underscoring in this setting. That problem lies at the heart of the debate

over the relationship between front line agencies, on the one hand, and OIRA, on the other. In the latest round of

this debate, Lisa Heinzerling, a former counsel and assistant administrator at the Environmental Protection

Agency, observes that generalist economists at OIRA fail to understand the nature of the problems line agencies

face.210 The OIRA reviewers, often newly minted Ph.D.s without training or experience in the substance of the

issues they confront, are prone to focus on the most easily observable and quantifiable components of the

problems, to the neglect of less-observable components that line officers know from long experience to be

critical.211

PDF GENERATED BY SEARCH.PROQUEST.COM Page 25 of 39

The tempting response to this problem is to simply call for vertical communication: whatever local complexities

line agencies perceive can simply be communicated upwards, and be folded into the process of regulatory

analysis.212 But things are not so simple. The OIRA reviewers, on the one hand, and the line agency officers, on

the other, deploy two different and potentially incompatible forms of knowledge. The former deploy explicit,

general, abstract, verbalized, and even quantified knowledge, the sort of knowledge obtained in a Ph.D. program.

The latter deploy local, experiential, how-to knowledge accumulated through long exposure to the specifics of

everevolving problems in the field at hand. The 30-year-old economist fresh from the University of Chicago will

sometimes be simply unable to evaluate, or perhaps even comprehend, the assumptions, intuitions, and

decisionmaking framework of the 60-year-old EPA bureau chief who has, perhaps, been involved with the problems

of coal ash or water-intake structures longer than the reviewer has been alive. Recall the earlier example-the

question whether automobile executives are bluffing about the capabilities of their firms and about the costs to

those firms of various regulatory outcomes. It is unclear whether such savvy can be explained, or how the OIRA

line reviewer is supposed to assess it.

On a Hayekian perspective, it is hopeless to think all relevant information can be conveyed by line agencies to a

centralized overseer like OIRA-a central planner for the executive branch.213 Agencies learn by doing,214 by

understanding the "circumstances of time and place."215 The tacit practical knowledge of line agencies is a form

of métis,216 knowing-how rather than knowing-that, and will by the nature of the case be inaccessible to experts in

OIRA-no matter how technically specialized-who have not themselves worked through the myriad complexities of

implementing general statutory commands and policies. No amount of meetings between line agencies and OIRA

personnel, or hiring of new OIRA personnel, will convey this tacit local knowledge to the center. Nor would it be

worth the costs in any event.

So the point about tacit expertise applies to OIRA as much as to the courts. On the other hand, one of our central

arguments does not apply to OIRA at all: the argument that courts may not require agencies to engage in

quantified cost-benefit analysis, even presumptively (unless of course organic statutes clearly so require). That

argument differs from the others on two counts: it is dependent on the substantive law, which is different for the

courts and for OIRA, and it is dependent on an institutional conception of the judicial role, which is obviously

sensitive to the identity of the reviewer. As for substantive law, the executive orders governing OIRA review of

executive branch agencies sometimes require (at least as a procedural matter) quantified cost-benefit analysis

insofar as feasible, unless special nonquantifiable factors are present.217 We have argued that there is no such

governing law for courts, however, and indeed that a global judicial requirement of (presumptive) quantified cost-

benefit analysis would be inconsistent with Congress's highly variable approach to decisionmaking requirements

in regulatory statutes.

As for the judicial role, we see quantified cost-benefit analysis as a particular, highly controversial decision-

procedure. Under Vermont Yankee and more recently Perez, courts have no authority to foist such a procedure on

agencies who otherwise possess legal discretion to adopt other decision-procedures, from the very large space of

possibilities.218 Clearly that point also does not apply to OIRA. As far as law goes, then, we have no quarrel with

quantified cost-benefit analysis internal to the executive branch. As for practical impact, however, it is also true

that an OIRA requirement of cost-benefit analysis, in a world where arbitrariness review does not itself require

costbenefit analysis for agencies, can sometimes create analytic and pragmatic awkwardness for agencies. In

some settings, it means that agencies will give reasons in cost-benefit format that are not the actual reasons for

the action, because the cost-benefit analysis was performed only to satisfy OIRA.219

Conclusion: The Baltimore Gas Era

State Farm and Chevron are said to be two of the pillars of administrative law. Many others before us have noted

that they are in some tension, with Chevron ushering in an era of deferential review of agency legal interpretation

and State Farm ushering in an era of robust judicial review of agency policymaking. The historical reality, however,

is actually quite the contrary. State Farm did not usher in an era of aggressive hard look review. In the Supreme

Court, agencies virtually never lose so-called hard look cases, and while the lower-court practice is more

PDF GENERATED BY SEARCH.PROQUEST.COM Page 26 of 39

heterogeneous, and includes highly intrusive outliers, State Farm review in practice is not systematically hard look.

It is time for the academic commentary to update.

As far as the law in action is concerned, we live in the era of Baltimore Gas review. Baltimore Gas made clear (1)

that it is generally sufficient that an agency states the nature of its uncertainty-not that it resolve it; (2) that

agencies are entitled to adopt any rational assumptions to cope with uncertainty, including highly optimistic

assumptions, which are just as rational as highly pessimistic ones; and (3) that courts may not demand the

impossible by requiring agencies to explain why they have chosen the assumptions they have, as opposed to other

assumptions.220 Baltimore Gas review is in fact more consistent with Supreme Court practice in the past three

decades than is State Farm (at least in its inflated form, as hard look review). When lower courts have strayed

toward a thick form of rationality review, the Court has been quick to overrule.

Rightly understood, arbitrary and capricious review is thin. It does not require agencies to use cost-benefit

analysis; it does not require the resolution of scientific uncertainty; it does not require that agencies pick the

optimal policy, or the most accurate policy, or the best feasible policy, or anything of that sort. It simply requires

that agencies act based on reasons. (As we have noted, there is a separate question whether agencies should be

obliged to give reasons for their actions). The set of admissible reasons includes second-order or higher-order

reasons for acting nonrationally or arbitrarily, as opposed to fully specified first-order reasons. Does this mean the

end of judicial review of agency decisionmaking? Not in the slightest. The Administrative Procedure Act says that

agency action may not be arbitrary and capricious, and we have proposed a straightforward interpretation of that

command, requiring reasoned decisionmaking-an interpretation that is not larded with all the fat of current

doctrine, and is thus more faithful to the Act's text.

Commentators often refer to arbitrariness, but no one ever discusses capriciousness. Perhaps caprice is a useful

lens for understanding the narrow set of circumstances in which agency action should be set aside under

706(2)(A). On our account, an agency's decision must be upheld unless it is based on nothing more than caprice, "a

sudden, impulsive, and seemingly unmotivated notion or action."221

Appendix

Our aggregate statistics comprise a list of all Supreme Court merits decisions involving an arbitrary and capricious

holding from 1983-2014. There are 65 cases total in our data, including State Farm itself. Four of the cases were

technically decided before State Farm, but during the same Term (October Term 1982); an example is Baltimore

Gas. The case list was generated by multiple independent Westlaw searches of "arbitrary and capricious" in the

Supreme Court database(s). From that initial return, we exclude any habeas or criminal cases that did not involve

an agency action and any pure constitutional challenges that did not involve an agency action.

Coding most of these cases is relatively straightforward. If the agency wins, the action was not arbitrary and

capricious. If the agency loses, it was arbitrary and capricious. The agency win-rate, so to speak, is simply the

number of wins divided by the total number of cases (wins and losses). Cases like Judulang v. Holder, in which

Chevron Step Two analysis fuses with arbitrariness review,222 are coded as simple losses for the agency on

arbitrariness grounds, thereby biasing the count against our thesis.

A minor difficulty arises in the handful of cases in which an opinion relies on both statutory and arbitrariness

analysis, in the alternative, to rule against the agency. In these cases, there are three methodological options: (1)

count the case as arbitrary in the numerator and include it in the denominator; (2) count the case as not arbitrary in

the numerator and count it in the denominator; or (3) count the case neither in the numerator nor the denominator.

The first is the most conservative estimate for our analysis; the second, the most forgiving.

These three approaches produce aggregate arbitrariness loss-rates of (1) 8/64=0.125; (2) 5/64=0.078; (3)

8/61=0.131. This choice, then produces a range of agency win-rates between 87 percent and 92 percent. The

number of debatable coding decisions is so small as to make no difference. However specified, our basic point

remains: agencies win the overwhelming majority of arbitrariness challenges in the Supreme Court. When a case

clearly involves a mix of statutory and arbitrariness analysis and the agency loses, we note as much in the case

list.

PDF GENERATED BY SEARCH.PROQUEST.COM Page 27 of 39

Footnote

1. 5 U.S.C. §706(2)(A) (2012).

2. White Stallion Energy Ctr., LLC v. EPA, 748 F.3d 1222, 1233 (D.C. Cir. 2014) (quoting Nat. Res. Def. Council, Inc. v.

EPA, 194 F.3d 130, 136 (D.C. Cir. 1999)).

3. See Pac. States Box &Basket Co. v. White, 296 U.S. 176, 181-82 (1935).

4. 401 U.S. 402, 416-17 (1971).

5. 463 U.S. 29, 42-43 (1983).

6. See State Farm, 463 U.S at 43 n.9 ("We do not view as equivalent the presumption of constitutionality afforded

legislation drafted by Congress and the presumption of regularity afforded an agency in fulfilling its statutory

mandate.").

7. See Louis J. Virelli III, Deconstructing Arbitrary and Capricious Review, 92 N.C. L. Rev. 721, 737-39 (2014).

8. See infra Part II.

9. This reads "arbitrary" in a decision-theory sense rather than a legal sense. Our argument is precisely that where

agencies have valid second-order reasons for acting inaccurately, nonrationally, or arbitrarily, they should not be

deemed to have acted in an "arbitrary and capricious" manner within the meaning of Section 706(2)(A) of the APA.

See Adrian Vermeule, Rationally Arbitrary Decisions in Administrative Law, 44 J. Legal Stud. S475, S482-89 (2015).

10. See infra Section II.A.

11. We put aside the limited cases, in the federal system, in which specialized agencies are reviewed by

specialized Article III courts. Most federal agency action that is reviewed by any Article III court is reviewed by a

general-purpose Article III court. See generally James E. Pfander, Article I Tribunals, Article III Courts, and the

Judicial Power of the United States, 118 Harv. L. Rev. 643 (2004).We also put aside review by specialized Article I

courts.

12. See, e.g., infra text accompanying notes 35-39 (discussing Judulang v. Holder, 132 S. Ct. 476 (2011)).

13. See infra Part I.

14. See infra Appendix.

15. See infra Appendix. This aggregate statistic depends on the classification of some cases as either statutory or

arbitrariness holdings. A few cases are blurry because the Court rules against the agency on alternative grounds,

both statutory interpretation and arbitrariness review. In very few cases, there is ongoing confusion about the so-

called second step of analysis under Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837

(1984); these decisions fold arbitrariness analysis into Chevron Step Two. See, e.g., Judulang v. Holder, 132 S. Ct.

476, 483 n.7 (2011) (stating that Chevron Step Two examines whether the agency interpretation is "arbitrary or

capricious in substance" (quoting Mayo Found. for Med. Educ. &Research v. United States, 562 U.S. 44, 53

(2011))). However, there are not enough debatable cases of either type to make any real difference; even on the

most conservative possible estimate (the estimate maximally biased against our thesis), agencies win

arbitrariness challenges in the Supreme Court about 88 percent of the time. We provide further details in the

Appendix. The overall trend in the data is clear: it is very rare indeed for the Supreme Court to hold an agency

action arbitrary and capricious.

16. 462 U.S. 87 (1983).

17. 556 U.S. 502 (2009).

18. 134 S. Ct. 1584 (2014).

19. 132 S. Ct. 476 (2011).

20. Balt. Gas &Elec. Co., 462 U.S. at 87-88.

21. Id. at 95.

22. Id. at 97 (alterations in original) (quoting Vt. Yankee Nuclear Power Corp. v. Nat. Res. Def. Council, Inc., 435 U.S.

519, 558 (1978)).

23. Id. at 103.

24. See, e.g., NLRB v. Noel Canning, 134 S. Ct. 2550, 2559-60 (2014); Zivotofsky v. Clinton, 132 S. Ct. 1421, 1427-28

PDF GENERATED BY SEARCH.PROQUEST.COM Page 28 of 39

(2012) ("At least since Marbury v. Madison . . . we have recognized that when an Act of Congress is alleged to

conflict with the Constitution, '[i]t is emphatically the province and duty of the judicial department to say what the

law is.' " (citation omitted) (quoting Marbury v. Madison, 5 U.S. (1 Cranch) 137 (1803))).

25. See infra text accompanying notes 75-76.

26. See infra Appendix.

27. See infra Appendix.

28. 549 U.S. 497 (2007).

29. 554 U.S. 527 (2008).

30. Massachusetts v. EPA, 549 U.S. at 534-35.

31. For an overview of the law bearing on review of agency "inaction," see Cass R. Sunstein &Adrian Vermeule, The

Law of "Not Now": When Agencies Defer Decisions, 103 Geo. L.J. 157 (2014). Very roughly speaking, the Court says

that (1) agency nonenforcement decisions are presumptively unreviewable, Heckler v. Chaney, 470 U.S. 821 (1985);

(2) agency refusals to engage in rulemaking are reviewable, although that review is "extremely limited" and "highly

deferential," Massachusetts v. EPA, 549 U.S. at 527-28.

32. There is an exception to the presumption against nonreviewability of agency enforcement decisions for cases

in which the agency has entirely abdicated its enforcement responsibility. See Heckler, 470 U.S. at 833 n.4;

Sunstein &Vermeule, supra note 31.

33. 132 S. Ct. 476 (2011).

34. Judulang, 132 S. Ct. at 479.

35. See id. at 485.

36. Id.

37. Id.; see also Sharon B. Jacobs, The Administrative State's Passive Virtues, 66 Admin. L. Rev. 565, 615 (2014)

(discussing this aspect of the decision).

38. See infra Appendix.

39. See infra Appendix.

40. See infra Appendix.

41. See infra Appendix.

42. Thomas J. Miles &Cass R. Sunstein, The Real World of Arbitrariness Review, 75 U. Chi. L. Rev. 761, 765 (2008).

43. See Richard J. Pierce, Jr., What Do the Studies of Judicial Review of Agency Actions Mean?, 63 Admin. L. Rev.

77, 84 (2011).

44. Miles &Sunstein, supra note 42, at 767.

45. Id. at 766.

46. 653 cases were reviewed, with 554 from the NLRB and the remainder from the EPA. See id. at 774.

47. E.g., Stephen G. Breyer et al., Administrative Law and Regulatory Policy 521-22 (7th ed. 2011).

48. See Miles &Sunstein, supra note 42, at 777.

49. See id. at 778-79.

50. See id.

51. 647 F.3d 1144, 1146 (D.C. Cir. 2011).

52. Business Roundtable, 647 F.3d at 1150 (alteration in original) (citation omitted) (citing Pub. Citizen v. Fed.

Motor Carrier Safety Admin., 374 F.3d 1209, 1221 (D.C. Cir. 2004).

53. See, e.g., John C. Coates IV, Cost-Benefit Analysis of Financial Regulation: Case Studies and Implications, 124

Yale L.J. 882, 917-20, 917 n.116 (2015); James D. Cox &Benjamin J.C. Baucom, The Emperor Has No Clothes:

Confronting the D.C. Circuit's Usurpation of SEC Rulemaking Authority, 90 Tex. L. Rev. 1811, 1813 (2012); Jill E.

Fisch, The Long Road Back: Business Roundtable and the Future of SEC Rulemaking, 36 Seattle U. L. Rev. 695,

697-98 (2013); Grant M. Hayden &Matthew T. Bodie, The Bizarre Law and Economics of Business Roundtable v.

SEC, 38 J. Corp. L. 101, 102 (2012); Bruce Kraus &Connor Raso, Rational Boundaries for SEC Cost-Benefit Analysis,

30 Yale J. on Reg. 289, 315-17 (2013); Cass R. Sunstein &Adrian Vermeule, Libertarian Administrative Law, 82 U.

PDF GENERATED BY SEARCH.PROQUEST.COM Page 29 of 39

Chi. L. Rev. 393, 440-48 (2015).

54. Sunstein &Vermeule supra note 53, at 440-42.

55. For discussion and references, see id. at 437-48.

56. For distinctions among brute uncertainty, strategic uncertainty, and model uncertainty in administrative

policymaking, see Vermeule, supra note 9, at S9-S13.

57. See Jeffrey N. Gordon, The Empty Call for Benefit-Cost Analysis in Financial Regulation, 43 J. Legal Stud. S351

(2014).

58. For the details, see Sunstein &Vermeule, supra note 53, at 435-48.

59. Vt. Yankee Nuclear Power Corp. v. Nat. Res. Def. Council, Inc., 435 U.S. 519 (1978).

60. Balt. Gas &Elec. Co. v. Nat. Res. Def. Council, Inc., 462 U.S. 87 (1983).

61. 135 S. Ct. 1199 (2015).

62. See George L. Priest and Benjamin Klein, The Selection of Disputes for Litigation, 13 J. Legal Stud. 1 (1984).

63. See David Zaring, Reasonable Agencies, 96 Va. L. Rev. 135 (2010).

64. See id. at 169.

65. Id.

66. See Thomas O. McGarity, Some Thoughts on "Deossifying" the Rulemaking Process, 41 Duke L.J. 1385, 1385-

86 (1992); Richard J. Pierce, Jr., Seven Ways to Deossify Agency Rulemaking, 47 Admin. L. Rev. 59, 60-62 (1995);

Mark Seidenfeld, Demystifying Deossification: Rethinking Recent Proposals to Modify Judicial Review of Notice

and Comment Rulemaking, 75 Tex. L. Rev. 483, 483 n.1 (1997).

67. See Stephen M. Johnson, Ossification's Demise? An Empirical Analysis of EPA Rulemaking from 2001-2005, 38

Envtl. L. 767, 784 (2008).

68. Anne Joseph O'Connell, Political Cycles of Rulemaking: An Empirical Portrait of the Modern Administrative

State, 94 Va. L. Rev. 889, 932 (2008).

69. Jason Webb Yackee &Susan Webb Yackee, Administrative Procedures and Bureaucratic Performance: Is

Federal Rule-Making "Ossified"?, 20 J. Pub. Admin. Res. &Theory 261 (2009).

70. Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983) (quoting Burlington Truck

Lines, Inc. v. United States, 371 U.S. 156, 168) (1962)).

71. Citizens to Preserve Overton Park, Inc. v. Volpe, 401 U.S. 402, 416 (1971).

72. State Farm, 463 U.S. at 51 ("Nor do we broadly require an agency to consider all policy alternatives in reaching

decision. It is true that rulemaking 'cannot be found wanting simply because the agency failed to include every

alternative device and thought conceivable by the mind of man . . . regardless of how uncommon or unknown that

alternative may have been . . . .' But the airbag is more than a policy alternative to the passive restraint Standard; it

is a technological alternative within the ambit of the existing Standard.") (alterations in original) (quoting Vt.

Yankee Nuclear Power Corp. v. Nat. Res. Def. Council, 435 U.S. 519, 551 (1978)).

73. See Pension Benefit Guar. Corp. v. LTV Corp., 496 U.S. 633, 646 (1990) ("If agency action may be disturbed

whenever a reviewing court is able to point to an arguably relevant statutory policy that was not explicitly

considered, then a very large number of agency decisions might be open to judicial invalidation.").

74. See id. at 647-52.

75. See City of Arlington v. FCC, 133 S. Ct. 1863 (2013).

76. Chevron U.S.A. Inc. v. Natural Res. Def. Council, Inc., 467 U.S. 837, 843-45 (1984).

77. See Mobil Oil Expl. &Producing Se., Inc. v. United Distrib. Cos., 498 U.S. 211, 230-31 (1991) ("An agency enjoys

broad discretion in determining how best to handle related, yet discrete, issues in terms of procedures and

priorities. . . . [A]n agency need not solve every problem before it in the same proceeding. This applies even where

the initial solution to one problem had adverse consequences for another area that the agency was addressing."

(citations omitted)). There is some tension between this principle and the Court's recent holding in Michigan v.

EPA, that an agency instructed to consider all "appropriate and necessary" factors must consider costs and

benefits at every stage of the proceedings. See 135 S. Ct. 2699, 2709 (2015); cf. id. at 2714 (Kagan, J., dissenting)

PDF GENERATED BY SEARCH.PROQUEST.COM Page 30 of 39

("The Agency acted well within its authority in declining to consider costs at the opening bell of the regulatory

process given that it would do so in every round thereafter . . . ."). As we discuss later, however, Michigan v. EPA is

best understood to stand for the narrow, indisputable, and indeed nearly tautological proposition that at every

stage of the administrative process, an agency decisionmaker must always consider both the pros and cons of

whatever course of action the decisionmaker undertakes; it cannot look only to one side of the ledger. See infra

text accompanying notes 137-146. So read, Michigan v. EPA is entirely compatible with Mobil Oil and with thin

rationality review generally. After all, although the agency must consider the pros and cons of every decision it

makes, it still enjoys "broad discretion" over timing and resource allocation, and "need not solve every problem

before it in the same proceeding." See Mobil Oil, 498 U.S. at 230-31.

78. "[W]hether all firms in the industry should be dealt with in a single proceeding or should receive individualized

treatment are questions that call for discretionary determination by the administrative agency." Moog Indus., Inc.

v. FTC, 355 U.S. 411, 413-14 (1958) (allowing the FTC to cull one firm from the herd).

79. See Thomas W. Merrill &Kristin E. Hickman, Chevron's Domain, 89 Geo. L.J. 833, 873-89 (2001); Cass R.

Sunstein, Chevron Step Zero, 92 Va. L. Rev. 187, 207-19 (2006).

80. FDA v. Brown &Williamson Tobacco Corp. and King v. Burwell both say that Chevron deference does not apply

to questions of great "economic and political significance." King v. Burwell, 135 S. Ct. 2480, 2488-89 (2015); FDA v.

Brown &Williamson Tobacco Corp., 529 U.S. 120, 159-60 (2000).

81. Sunstein's separate section in Libertarian Administrative Law-with which Vermeule emphatically disagrees, see

Sunstein &Vermeule supra note 53, at 443-46- seemingly endorses the following propositions:

To the extent that available evidence permits quantification, it would be arbitrary not to quantify. . . . To the extent

that Business Roundtable stands for this general principle, it is on firm ground. . . . Indeed, it would generally seem

arbitrary for an agency to issue a rule that has net costs (or no net benefits), at least unless a statute requires it to

do so.

Id. at 441. For further discussion, see also Justice Scalia's statement at oral argument in Michigan v. EPA:

I'm not even sure I agree with the premise that when . . . Congress says nothing about cost, the agency is entitled

to disregard cost. I would think it's classic arbitrary and capricious agency action for an agency to command

something that is outrageously expensive and . . . in which the expense vastly exceeds whatever public benefit can

be . . . achieved. I would think that's . . . a violation of the Administrative Procedure Act.

Transcript of Oral Argument at 14, Michigan v. EPA, 135 S. Ct. 2699 (2015) (No. 14-46). The actual decision in

Michigan v. EPA did not go nearly that far, however. See infra text accompanying notes 137-148.

82. Compare Cass R. Sunstein, Risk and Reason: Safety, Law, and the Environment 6, 99-132 (2002), and Cass R.

Sunstein, Cognition and Cost-Benefit Analysis, 29 J. Legal Stud. 1059 (2000) (advocating cost-benefit analysis as a

corrective to cognitive problems and judgment errors), with Richard A. Posner, Catastrophe: Risk and Response

139 (2004), Sidney A. Shapiro &Christopher H. Schroeder, Beyond Cost-Benefit Analysis: A Pragmatic

Reorientation, 32 Harv. Envtl. L. Rev. 433, 446-50 (2008) (describing adoption of costbenefit analysis as part of a "

'comprehensive rationality' that rational choice theorists consider essential to rational decisionmaking"), and

Amartya Sen, The Discipline of Cost-Benefit Analysis, 29 J. Legal Stud. 931, 935 (2000) ("At the risk of

oversimplification, explicit valuation is a part of the insistence on a rationalist approach, which demands full

explication of the reasons for taking a decision, rather than relying on an unreasoned conviction or on an implicitly

derived conclusion.").

83. See Cass R. Sunstein, The Cost-Benefit State: The Future of Regulatory Protection 59-60 (2002) [hereinafter

Sunstein, The Cost-Benefit State]; Cass R. Sunstein, Cost-Benefit Default Principles, 99 Mich. L. Rev. 1651, 1691-94

(2001) [hereinafter Sunstein, Cost-Benefit Default Principles].

84. See Sunstein, Cost-Benefit Default Principles, supra note 83, at 1686.

85. For the view that quantified cost-benefit analysis is best understood as a decisionprocedure, see Matthew D.

Adler &Eric A. Posner, Rethinking Cost-Benefit Analysis, 109 Yale L.J. 165, 167 (1999). For the view that it is a

sectarian decision-procedure, see id. ("Many law professors, economists, and philosophers believe that CBA does

PDF GENERATED BY SEARCH.PROQUEST.COM Page 31 of 39

not produce morally relevant information and should not be used in project evaluation.").

86. See Sen, supra note 82, at 934.

87. See generally Sen, supra note 82 (describing a "general social choice approach" to cost-benefit analysis). For a

clear-minded treatment of the distinctions among intuitive judgment, unquantified cost-benefit analysis and

quantified cost-benefit analysis, see Coates, supra note 53, at 892-93.

88. 2 The Correspondence of Charles Darwin 1837-1843 444 (Frederick Burkhardt et al. eds., 1986). Cases like

Darwin's list actually suggest that some decisions ought not to be made even on the basis of informal cost-benefit

analysis, although we need not establish that proposition here.

89. See Facilitating Shareholder Director Nominations, 75 Fed. Reg. 56,668, 56,753-71 (Sept. 16, 2010) (to be

codified at 17 C.F.R. pts. 200, 232, 240, and 249).

90. See Bus. Roundtable v. SEC, 647 F.3d 1144, 1149-51 (D.C. Cir. 2011).

91. See, e.g., Matthew D. Adler &Eric A. Posner, New Foundations of Cost-Benefit Analysis 100 (2006).

92. See, e.g., id.; Richard L. Revesz &Michael A. Livermore, Retaking Rationality: How Cost-Benefit Analysis Can

Better Protect the Environment and Our Health 12-13 (2008); Sunstein, The Cost-Benefit State, supra note 83, at 6-

10.

93. See, e.g., Frank Ackerman &Lisa Heinzerling, Priceless: On Knowing the Price of Everything and the Value of

Nothing 7-12 (2004); Thomas O. McGarity, Reinventing Rationality: The Role of Regulatory Analysis in the Federal

Bureaucracy 142-59 (2005); Coates, supra note 53; Amy Sinden et al., Cost-Benefit Analysis: New Foundations on

Shifting Sand, 3 Reg. &Governance 48, 55-56 (2009).

94. See Adler &Posner, supra note 85, at 171-72 (discussing objections).

95. See id.

96. Confusingly, in recent years, some of the academic commentary has concluded that Chevron (or at least

Chevron Step Two) and arbitrariness review are actually equivalent. If an agency has adopted a view that is an

unreasonable interpretation of its organic statute, that view must also be arbitrary and capricious; if an agency

view is arbitrary and capricious then it cannot be a reasonable interpretation of the statute. See, e.g., Gary S.

Lawson, Reconceptualizing Chevron and Discretion: A Comment on Levin and Rubin, 72 Chi.-Kent L. Rev. 1377,

1377-79 (1997); Ronald M. Levin, The Anatomy of Chevron: Step Two Reconsidered, 72 Chi.Kent L. Rev. 1253, 1276

(1997); Laurence H. Silberman, Chevron-The Intersection of Law &Policy, 58 Geo. Wash. L. Rev. 821, 827-28 (1990).

For our purposes, this debate is something of a sideshow. We simply note it because the language of cases and

commentary is often imprecise, slipping back and forth between doctrinal frameworks for this reason.

97. See infra text accompanying note 217.

98. 435 U.S. 519 (1978).

99. 135 S. Ct. 1199, 1207 (2015).

100. See Michigan v. EPA, 135 S. Ct. 2699, 2708 (2015).

101. Perez, 135 S. Ct. at 1207. Michigan v. EPA clearly declined to impose any such requirement. 135 S. Ct. at

2711.

102. See Jonathan S. Masur &Eric A. Posner, Against Feasibility Analysis, 77 U. Chi. L. Rev. 657, 658 (2010)

(collecting examples). Masur and Posner criticize feasibility but do not claim that Congress should be taken to

intend to require that agencies perform quantified CBA whenever statutes are silent or ambiguous. Id. at 662-63,

687-712.

103. Wong Yang Sung v. McGrath, 339 U.S. 33, 40 (1950).

104. Cf. Cass R. Sunstein, Is OSHA Unconstitutional?, 94 Va. L. Rev. 1407 (2008). Sunstein criticizes OSHA on

nondelegation grounds but is also sharply critical of feasibility analysis. The logic of his view is that the Act is

patently irrational and therefore vulnerable to a due process challenge.

105. See Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 n.9 (1983) ("We do not view as

equivalent the presumption of constitutionality afforded legislation drafted by Congress and the presumption of

regularity afforded an agency in fulfilling its statutory mandate.").

PDF GENERATED BY SEARCH.PROQUEST.COM Page 32 of 39

106. See Amy Sinden, Cass Sunstein's Cost-Benefit Lite: Economics for Liberals, 29 Colum. J. Envtl. L. 191, 240

(2004); cf. Thomas O. McGarity, Professor Sunstein's Fuzzy Math, 90 Geo. L. J. 2341, 2343 (2002).

107. 556 U.S. 208 (2009).

108. Riverkeeper, Inc. v. EPA, 475 F.3d 83, 99-100 (2d Cir. 2007).

109. 33 U.S.C. §1326(b) (2012) (emphasis added).

110. Entergy, 556 U.S. at 215-16; National Pollutant Discharge Elimination System-Final Regulations to Establish

Requirements for Cooling Water Intake Structures at Phase II Existing Facilities, 69 Fed. Reg. 41,576, 41,599-606

(July 9, 2004) (to be codified at 40 C.F.R. pt. 23.2).

111. Entergy, 556 U.S. at 216.

112. Id. at 218.

113. Id.

114. Id.

115. Id. at 223 (discussing Whitman v. American Trucking Ass'ns., Inc., 531 U.S. 457 (2001)).

116. Id. (discussing Am. Textile Mfrs. Inst., Inc. v. Donovan, 452 U.S. 490 (1981)).

117. Id.

118. 134 S. Ct. 1584 (2014).

119. Clean Air Act, 42 U.S.C. §7410(a)(2)(D)(i) (2012).

120. EME Homer City Generation, L.P. v. EPA, 696 F.3d 7, 20-21 (D.C. Cir. 2012), rev'd, 134 S. Ct. 1584 (2014).

121. See Federal Implementation Plans: Interstate Transport of Fine Particulate Matter and Ozone and Correction

of SIP Approvals, 76 Fed. Reg. 48,208, 48,254 (Aug. 8, 2011) (to be codified at 40 C.F.R. pts. 51, 52, 72 et al.); see

also EME Homer City, 134 S. Ct. at 1597.

122. EME Homer City, 132 S. Ct. at 1607.

123. 134 S. Ct. 2427, 2434 (2014).

124. 549 U.S. 497, 528-29 (2007).

125. See UARG, 134 S. Ct. at 2436-37.

126. See Clean Air Act, 42 U.S.C. §§7471-7492 (2012).

127. Id. §7661.

128. See id. §§7479(1), 7661(2)(B), 7602(j).

129. See Regulating Greenhouse Gas Emissions Under the Clean Air Act, 73 Fed. Reg. 44,354, 44,498-99 (proposed

July 30, 2008); see also UARG, 134 S. Ct. at 2436.

130. See Prevention of Significant Deterioration and Title V Greenhouse Gas Tailoring Rule, 75 Fed. Reg. 31,514,

31,523-24 (June 3, 2010) (to be codified at 40 C.F.R. pts. 50-52, et al.).

131. UARG, 134 S. Ct. at 2443-46.

132. Id. at 2447-49.

133. Id. at 2447.

134. Id. at 2448.

135. Id. at 2441 (quoting Envtl. Def. v. Duke Energy Corp., 549 U.S. 561, 574 (2007)).

136. Id. at 2448.

137. 135 S. Ct. 2699 (2015).

138. Clean Air Act, 42 U.S.C. §7412 (2012).

139. Michigan v. EPA, 135 S. Ct. at 2706-09.

140. Id. at 2711-12.

141. Or straightforward for those who believe that Chevron has two steps. But see Matthew C. Stephenson &Adrian

Vermeule, Chevron Has Only One Step, 95 Va. L. Rev. 597 (2009).

142. E.g., Cass R. Sunstein, Thanks, Justice Scalia, for the Cost-Benefit State, Bloomberg View (July 7, 2015),

http://www.bloombergview.com/articles/2015-07-07/thanks-justicescalia-for-the-cost-benefit-state

[http://perma.cc/8ENH-NJWP] (calling the opinion in Michigan v. EPA a "ringing endorsement of cost-benefit

PDF GENERATED BY SEARCH.PROQUEST.COM Page 33 of 39

analysis by government agencies").

143. Michigan v. EPA, 135 S. Ct. at 2707.

144. Id. (emphasis omitted).

145. Id. at 2711; accord id. at 2717 (Kagan, J., dissenting) ("As the Court notes, [accounting for costs] does not

require an agency to conduct a formal cost-benefit analysis of every administrative action.").

146. Id. at 2707. This is not to say, of course, that Michigan v. EPA is correct. In our view, the Court slipped from an

unexceptional premise, that agencies should consider both the advantages and disadvantages of their decisions,

to the very different and indefensible conclusion that agencies must consider those things all together, at every

stage of regulatory proceedings. On the contrary, the background presumption of administrative law is that

agencies may parcel out the consideration of relevant factors into different stages of proceedings or even different

proceedings. See supra note 77 (discussing Mobil Oil Expl. &Producing Se., Inc. v. United Distrib. Cos., 498 U.S.

211, 230-31 (1991)). Although the presumption may of course be overcome by clear statutory instructions to the

contrary, the textual phrase "appropriate and necessary" should have been deemed insufficient to do so. See

Michigan v. EPA, 135 S. Ct. at 2714-15 (Kagan, J., dissenting) ("[The majority's] micromanagement of EPA's

rulemaking, based on little more than the word 'appropriate'-runs counter to Congress's allocation of authority

between the Agency and the courts.").

147. Catherine M. Sharkey, State Farm "With Teeth": Heightened Judicial Review in the Absence of Executive

Oversight, 89 N.Y.U. L. Rev. 1589, 1592 (2014).

148. See U.S. Envtl. Prot. Agency, EPA-SAB-EC-WKSHP-02-001, Workshop on the Benefits of Reductions in

Exposure to Hazardous Air Pollutants: Developing Best Estimates of Dose-Response Functions 2 (2002),

https://yosemite.epa.gov/sab%5CSAB

PRODUCT.NSF/34355712EC011A358525719A005BF6F6/$File/ecwkshp02001%2Bappa-g.pdf

[https://perma.cc/QHK2-YBML].

149. Id. at C-1.

150. Motor Vehicle Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 43 (1983) (quoting Burlington Truck

Lines v. United States, 371 U.S. 156, 168 (1962)).

151. See Frank H. Knight, Risk, Uncertainty and Profit (1921).

152. Jon Elster, Explaining Technical Change 199 (1983).

153. See id. at 185-207; Jon Elster, Nuts and Bolts for the Social Sciences 34-35 (1989); Adrian Vermeule, Judging

Under Uncertainty 171-72 (2006); David Kelsey &John Quiggin, Theories of Choice Under Ignorance and

Uncertainty, 6 J. Econ. Surveys 133 (1992).

154. See Kenneth J. Arrow &Leonid Hurwicz, An Optimality Criterion for Decision-making Under Ignorance, in

Uncertainty and Expectations in Economics 1 (C. F. Carter &J. L. Ford eds., 1972). The generalized version is the

so-called "a-maximin framework," which models decisionmakers as choosing a parameter that may range from

maximal pessimism to maximal optimism. For a review of the literature, with legal and regulatory applications, see

Daniel A. Farber, Uncertainty, 99 Geo. L.J. 901 (2011).

155. See Robertson v. Methow Valley Citizens Council, 490 U.S. 332, 356 (1989) (finding that under NEPA, court of

appeals erred in requiring agencies to perform a "worst case analysis" under uncertainty, in part because worst-

case analysis "distort[s] the decisionmaking process by overemphasizing highly speculative harms"); Balt. Gas

&Elec. Co. v. Nat. Res. Def. Council, Inc., 462 U.S. 87, 103 (1983) (upholding NRC's maximally optimistic "zero-

release" assumption for spent nuclear fuels, observing that "a reviewing court must generally be at its most

deferential" when an agency "mak[es] predictions, within its area of special expertise, at the frontiers of science").

156. See Vermeule, supra note 9.

157. Tucson Herpetological Soc'y v. Salazar, 566 F.3d 870 (9th Cir. 2009); Vermeule, supra note 9, at S9-S11.

158. Richard Murphy, We Found Out That Counting Lizard Poop Is Not a Good Way to Count Lizards: Now What?,

JOTWELL (Jan. 6, 2015) (emphasis added), http://adlaw.jotwell .com/we-found-out-that-counting-lizard-poop-is-

not-a-good-way-to-count-lizards-nowwhat/ [http://perma.cc/VH9D-LJDN].

PDF GENERATED BY SEARCH.PROQUEST.COM Page 34 of 39

159. See Massachusetts v. EPA, 549 U.S. 497 (2007).

160. See Tri-Valley CAREs v. U.S. Dep't of Energy, 671 F.3d 1113 (9th Cir. 2012); Jifry v. FAA, 370 F.3d 1174 (D.C.

Cir. 2004).

161. See Tucson Herpetological Soc'y, 566 F.3d 870.

162. See Bus. Roundtable v. SEC, 647 F.3d 1144 (D.C. Cir. 2011).

163. Vermeule, supra note 9, at S4.

164. Id.

165. Id.

166. Id. at S21. See generally Hans O. Melberg, A Critical Discussion of Jon Elster's Arguments About Rational

Choice, Infinite Regress and the Collection of Information (May 1999) (unpublished Ph.D. dissertation, University of

Oslo) (on file with author).

167. Jon Elster, Bad Timing, in The Thief of Time 87, 96 (Chrisoula Andreou &Mark D. White eds., 2010) (quoting

Leif Johansen, 1 Lectures on Macroeconomic Planning 144 (1977)).

168. See supra text accompanying notes 71-80.

169. Michael Byron, Satisficing and Optimality, 109 Ethics 67, 72 (1998).

170. See Michael Slote, Beyond Optimizing 7-10 (1989).

171. See Herbert A. Simon, Reason in Human Affairs 85 (1983); Byron, supra note 169, at 71; Herbert A. Simon, A

Behavioral Model of Rational Choice, 69 Q.J. Econ. 99, 112 (1955).

172. Cf. David Schmidtz, Rational Choice and Moral Agency 28-40 (1995) [hereinafter Schmidtz, Rational Choice];

David Schmidtz, Satisficing as a Humanly Rational Strategy, in Satisficing and Maximizing: Moral Theorists on

Practical Reason 30, 31 (Michael Byron ed., 2004).

173. In the presence of information costs and search costs, another strategy is constrained optimization. On this

approach, decisionmakers should invest in gathering information just up to the point at which the (increasing)

marginal costs of doing so equal the expected marginal benefits of further information. Cf. George J. Stigler, The

Economics of Information, 69 J. Pol. Econ. 213, 216 (1961). Whereas the satisficer stops when the choice at hand

is good enough, the constrained optimizer stops looking for a better choice when the marginal benefit of finding a

better option, discounted by the probability of finding such an option, is equal to or less than the marginal costs of

further search. The optimizing approach, however, assumes that the decisionmaker always has epistemically well-

grounded probability distributions over the marginal costs and benefits of further search-an assumption we reject.

In any event, the two strategies are simply different. "An optimizing strategy places limits on how much we are

willing to invest in seeking alternatives. A satisficing strategy places limits on how much we insist on finding

before we quit that search and turn our attention to other matters." Schmidtz, Rational Choice, supra note 172, at

34-35; see also Jonathan Brodie Bendor et al., Satisficing: A 'Pretty Good' Heuristic, 9 B.E. J. Theoretical Econ. 1

(2009). Finally, even were it true that satisficing is just a form of optimization-under-constraints, our substantive

point in text would be unaffected. Whether agency decisionmaking under uncertainty is described as satisficing or

as constrained optimization, the substantive point is that comparative evaluation of policies is not always required

by rationality-and FCC v. Fox Television Stations, Inc., 556 U.S. 502, 515 (2009), holds that it is not required by

administrative law. See infra text accompanying notes 182-188.

174. See Stephen Breyer, Judicial Review of Questions of Law and Policy, 38 Admin. L. Rev. 363, 393 (1986); see

also Vt. Yankee Nuclear Power Corp. v. Nat. Res. Def. Council, Inc., 435 U.S. 519, 551 (1978); Nat'l Wildlife Fed'n v.

FERC, 912 F.2d 1471, 1484-85 (D.C. Cir. 1990) (per curiam) (finding that FERC need not consider an alternative

project that was not reasonable under statutory requirements); City of New York v. U.S. Dep't of Transp., 715 F.2d

732, 744 (2d Cir. 1983) (finding that DOT need not consider the barging of nuclear waste, as it was an

unreasonable alternative).

175. Motor Vehicles Mfrs. Ass'n v. State Farm Mut. Auto. Ins. Co., 463 U.S. 29, 51 (1983).

176. Breyer, supra note 174, at 393.

177. For a stellar recent exception to this generalization, see Center for Sustainable Economy v. Jewell, 779 F.3d

PDF GENERATED BY SEARCH.PROQUEST.COM Page 35 of 39

588, 610-12 (D.C. Cir. 2015) (allowing Interior Department to treat option value-the informational value of delaying a

decision-in qualitative rather than quantitative terms).

178. Balt. Gas &Elec. Co. v. Nat. Res. Def. Council, Inc., 462 U.S. 87, 104 (1983).

179. On this view, the NRC's assumption of "zero-release" of spent nuclear fuels should not be taken too literally. It

was not a prediction that not one iota of such fuels would ever be released. Rather it was a policy choice, one that

opted for maximax assumptions-highly optimistic assumptions-in the presence of severe uncertainty. Id. at 92.

180. 556 U.S. 502 (2009).

181. Fox, 556 U.S. at 515.

182. See Breyer, supra note 174.

183. Fox, 556 U.S. at 519.

184. Id. at 515.

185. Ronald M. Levin summarizes Fox as follows: "The Court's position on the issue comes down to this: Unless

reliance interests or inconsistent readings of a factual record are involved, open acknowledgment of the change

and a defense of the new policy on its own terms should ordinarily suffice." Ronald M. Levin, Hard Look Review,

Policy Change, and Fox Television, 65 U. Miami L. Rev. 555, 573 (2011). The latter clause is what the Fox majority

and dissent disagreed about.

186. Mobil Oil Expl. &Producing Se., Inc. v. United Distrib. Cos., 498 U.S. 211, 230 (1991).

187. See discussion supra Section II.A.2.c.

188. We use the plus or minus nomenclature only for expositional purposes. The agency is summarizing a

probability distribution and we could explain with the mean-variance problem with greater formality. The point here

is a simple one, however.

189. See Jacob E. Gersen, Administrative Law Goes to Wall Street: The New Administrative Process, 65 Admin. L.

Rev. 689, 692, 724-25 (2013).

190. See Ctr. for Food Safety v. Hamburg, 954 F. Supp. 2d 965 (N.D. Cal. 2013).

191. See, e.g., Stephen Breyer, Breaking the Vicious Circle: Toward Effective Risk Regulation (1993) (explaining the

"vicious circle" caused by systemic problems with the regulatory system); McGarity, supra note 66 (criticizing the

rigid restraints imposed on informal rulemakings); Pierce, supra note 66 (criticizing judicially-induced rulemaking

ossification and discussing potential judicial solutions); Paul R. Verkuil, Commentary, Rulemaking Ossification-A

Modest Proposal, 47 Admin. L. Rev. 453 (1995) (criticizing Pierce's proposed judicial solutions to rulemaking

ossification and proposing alternative legislative solutions).

192. See, e.g., Jifry v. FAA, 370 F.3d 1174 (D.C. Cir. 2004) (upholding, against arbitrariness challenge, an FAA

regulation promulgated without notice and comment under the "good cause" exception); see also Adrian Vermeule,

Our Schmittian Administrative Law, 122 Harv. L. Rev. 1095, 1122-25 (2009) (examining the "good cause"

exception).

193. Louis J. Virelli III, Deconstructing Arbitrary and Capricious Review, 92 N.C. L. Rev. 721, 758-60 (2014).

194. Daniel Epps, The Consequences of Error in Criminal Justice, 128 Harv. L. Rev. 1065 (2015) (describing and

challenging the Blackstone principle).

195. See, e.g., id. at 1094-124 (offering a "dynamic" critique of the Blackstone principle based on its systemic

consequences for the criminal justice system).

196. See Barnabas Dickson, The Precautionary Principle in CITES: A Critical Assessment, 39 Nat. Resources J. 211,

211-15 (1999).

197. For an overview of relevant philosophical literature, see Harry Collins, Tacit and Explicit Knowledge (2010).

198. F. A. Hayek, The Use of Knowledge in Society, 35 Am. Econ. Rev. 519, 524 (1945).

199. For an illuminating analysis of "craft expertise" based on experience, very much compatible with our approach

here, see Sidney A. Shapiro, The Failure to Understand Expertise in Administrative Law: The Problem and the

Consequences (Jan. 7, 2016) (unpublished manuscript) (on file with Michigan Law Review). For a first stab at the

"tacit expertise" issue, see Adrian Vermeule, Local and Global Knowledge in the Administrative State, in Law,

PDF GENERATED BY SEARCH.PROQUEST.COM Page 36 of 39

Liberty and State: Oakeshott, Hayek and Schmitt on the Rule of Law 295 (David Dyzenhaus &Thomas Poole eds.,

2015).

200. Office of Mgmt. &Budget, OMB Circular A-4, Regulatory Analysis 2-3 (2003). For a comprehensive and

persuasive defense of professional judgment in regulatory analysis, see Coates, supra note 53.

201. Cf. Jennifer Nou, Agency Self-Insulation Under Presidential Review, 126 Harv. L. Rev. 1755 (2013) (exploring

how agencies may act strategically to avoid presidential review).

202. See supra Section II.A.5.c.

203. See supra text accompanying note 198.

204. Donald MacKenzie &Graham Spinardi, Tacit Knowledge and the Uninvention of Nuclear Weapons, in Knowing

Machines: Essays on Technical Change 215, 256 (Donald MacKenzie ed., 1996).

205. Donald MacKenzie, Moving Toward Disinvention, Bull. Atomic Scientists, Sept.-Oct. 1996, at 4, 4:

Testing made designers' competence visible. In heated design reviews at the labs, those whose opinions really

counted were the "test-seasoned" designers. They had shown that they knew when computer models could be

relied upon, and when they could not. So they were fit people to judge neophytes' proposals. . . . [Veteran

designers] warned us repeatedly of the dangers of reliance on simulations and laboratory tests alone. "You start to

believe your calculations, and young folks really believe them if the old timers have left," said one. "People start to

believe the codes are absolutely true, to lose touch with reality," said another. . . . One key issue is the distinction

between explicit and tacit knowledge. Explicit knowledge is knowledge that can be written down and preserved in

diagrams, documents, and computer files. Tacit knowledge is the motor skills, intuition, "common sense," and

judgment that cannot be transmitted in words or equations alone. Both are vital in scientific and technical

endeavor. Few of us would wish to come under the knife of a surgeon who possessed only explicit knowledge. The

history of nuclear weaponry is of repeated discovery that explicit knowledge alone is not enough. . . . The ultimate

value of nuclear testing to weapons designers was that it served as a check of the quality of the tacit knowledge

they, and those who built the weapons, brought to their task. . . . Knowledge of the physics that makes nuclear

weapons possible is humanity's permanent inheritance. But that is not true of the assembled, partially tacit, largely

engineering-based skills that make nuclear weapons technological realities rather than drawing-board concepts.

See also MacKenzie &Spinardi, supra note 204, at 215-58.

206. McGarity, supra note 93, at 281.

207. This example was helpfully suggested by Lisa Heinzerling (personal communication).

208. See Miles &Sunstein, supra note 42, at 796-97.

209. See supra text accompanying notes 3-7.

210. See Lisa Heinzerling, Inside EPA: A Former Insider's Reflections on the Relationship Between the Obama EPA

and the Obama White House, 31 Pace Envtl. L. Rev. 325, 355-56 (2014).

211. This is a well-known problem in contract theory as well. When there are multiple tasks, agents may expend

too much effort on the readily observable task precisely because that is the only observable action on which

sanction and reward can be conditioned. See Bengt Holmstrom &Paul Milgrom, Multitask Principal-Agent

Analyses: Incentive Contracts, Asset Ownership, and Job Design, 7 J.L. Econ. &Org. (Special Issue) 24, 27-28

(1991).

212. Cf. Cass R. Sunstein, The Office of Information and Regulatory Affairs: Myths and Realities, 126 Harv. L. Rev.

1838, 1840-42, 1854-56 (2013) (arguing that OIRA should be elevated because of its information-sharing capacity).

213. See Vermeuele, supra note 199, at 297.

214. Cf. Sean Gailmard &John W. Patty, Learning While Governing: Expertise and Accountability in the Executive

Branch (2012) (discussing expertise development and information acquisition in the executive branch).

215. See Hayek, supra note 198, at 521.

216. See James C. Scott, Seeing Like a State 309-41 (1998).

217. By their own terms, these executive orders are not judicially enforceable. See Exec. Order No. 12,866, 3 C.F.R.

638, 649 (1994), reprinted in 5 U.S.C. §601 (2012)

PDF GENERATED BY SEARCH.PROQUEST.COM Page 37 of 39

Nothing in this Executive order shall affect any otherwise available judicial review of agency action. This Executive

order is intended only to improve the internal management of the Federal Government and does not create any

right or benefit, substantive or procedural, enforceable at law or equity by a party against the United States, its

agencies or instrumentalities, its officers or employees, or any other person.

Exec. Order No. 13,563, 3 C.F.R. 215, 217 (2012), reprinted in 5 U.S.C. §601 (2012) (same); see also Meyer v. Bush,

981 F.2d 1288, 1296 n.8 (D.C. Cir. 1993) ("An Executive Order devoted solely to the internal management of the

executive branch-and one which does not create any private rights-is not . . . subject to judicial review."); Michigan

v. Thomas, 805 F.2d 176, 187 (6th Cir. 1986) ("The Order was intended 'to improve the internal management of the

Federal government' and not to confer rights judicially enforceable in private litigation." (quoting Exec. Order No.

12,291, 3 C.F.R. 127 (1982))).

218. Perez v. Mortg. Bankers Ass'n, 135 S. Ct. 1199, 1207 (2015); Vt. Yankee Nuclear Power Corp. v. Nat. Res. Def.

Council, 435 U.S. 519, 524, 543-45 (1978).

219. See Heinzerling, supra note 210, at 349-58.

220. See Balt. Gas &Elec. Co. v. Nat. Res. Def. Council, 462 U.S. 87, 98-106 (1983).

221. Caprice, Merriam-Webster, http://www.merriam-webster.com/dictionary/caprice [h8ttp://perma.cc/UU4K-

GFEU].

222. See Judulang v. Holder, 132 S. Ct. 476, 483 n.7 (2011) (stating that Chevron Step Two examines whether the

agency interpretation is "arbitrary or capricious in substance" (quoting Mayo Found. for Med. Educ. &Research v.

United States, 562 U.S. 44, 53 (2011))). For discussion and citations of the episodic fusion of Chevron Step Two

with arbitrariness review, see Stephenson &Vermeule, supra note 141.

AuthorAffiliation

Jacob Gersen* and Adrian Vermeule**

* Professor of Law and Affiliated Professor of Government, Harvard Law School and Department of Government.

** John H. Watson, Jr. Professor of Law, Harvard Law School. For helpful comments, we are very grateful to Judge

Harry T. Edwards, Nick Parillo, David Strauss and to participants at a University of Chicago conference on "Federal

Agency Decision Making Under Deep Uncertainty," May 8-9, 2015. Thanks to Evelyn Blacklock, Cindy Guan, and

Rauvin Johl for outstanding research assistance. DETAILS

Subject: Rationality; Federal court decisions; Judges &magistrates; Judicial reviews

Location: United States--US

Classification: 4330: Litigation; 9190: United States

Publication title: Michigan Law Review; Ann Arbor

Volume: 114

Issue: 8

Pages: 1355-1412

Number of pages: 58

Publication year: 2016

PDF GENERATED BY SEARCH.PROQUEST.COM Page 38 of 39

Database copyright  2019 ProQuest LLC. All rights reserved. Terms and Conditions Contact ProQuest

Publication date: Jun 2016

Publisher: Michigan Law Review Association

Place of publication: Ann Arbor

Country of publication: United States, Ann Arbor

Publication subject: Law

ISSN: 00262234

Source type: Scholarly Journals

Language of publication: English

Document type: Feature

Document feature: References

ProQuest document ID: 1794161482

Document URL: http://0-

search.proquest.com.wizard.umd.umich.edu/docview/1794161482?accountid=1457

8

Copyright: Copyright Michigan Law Review Association Jun 2016

Last updated: 2016-06-21

Database: ABI/INFORM Collection,Research Library

PDF GENERATED BY SEARCH.PROQUEST.COM Page 39 of 39