ISSN 1045-6333

HARVARD JOHN M. OLIN CENTER FOR LAW, ECONOMICS AND BUSINESS

PREFERENCE-ESTIMATINGSTATUTORY DEFAULT RULES

Einer Elhauge

Discussion Paper No. 378

07/2002

Harvard Law SchoolCambridge, MA 02138

The Center for Law, Economics, and Business is supported bya grant from the John M. Olin Foundation.

This paper can be downloaded without charge from:

The Harvard John M. Olin Discussion Paper Series:http://www.law.harvard.edu/programs/olin_center/

JEL Classifications: D70, D72, D73, D78, D99, H10, H80, K00, K20, K30, K40

PREFERENCE-ESTIMATING STATUTORY DEFAULT RULES

(forthcoming in Columbia Law Review)

By Einer Elhauge*

ABSTRACT

It is commonly assumed that statutory indeterminancy must be resolved by judicial judgment. ThisArticle argues that where hermeneutics gives out, statutory indeterminancy instead can, is, andshould be resolved by default rules designed to minimize the expected dissatisfaction of enactablepreferences. This probabilistic goal justifies many judicial practices such as broad-ranging inquiriesinto legislative history even if it does not accurately reveal any shared legislative intent. It alsooften supports adopting moderate interpretations even when more extreme interpretations are morelikely to match legislative preferences. Further, while the general default rule normally requiresestimating enacting legislative preferences, the enacting legislature itself would prefer to shift to adefault rule of tracking current legislative preferences when those can be reliably ascertained fromofficial action. The basic reason is that the enacting legislature would prefer influence over theinterpretation of the entire stock of statutes being interpreted while it is office rather than influenceover the future interpretation (when it is out of office) of the statutory ambiguities that exist on thefew topics for which it made enactments. Such a current preferences default rule explains manycases that rely on subsequent legislative action despite its hermeneutic irrelevance, and explainsboth the general doctrine of deference to agency interpretations and the pattern of exceptions tothat deference.

*Professor of Law, Harvard Law School. Email: elhauge@law.harvard.edu. I am grateful for funding from Harvard LawSchool and the John M. Olin Center for Law, Economics and Business at Harvard Law School. I have also benefittedfrom comments on earlier versions of this paper by Scott Altman, Ian Ayres, Sam Bagenstos, Jack Balkin, JosephBankman, David Barron, Jack Beerman, Yochai Benkler, Victory Brudney, Thomas Campbell, Alex Capron, CaryCoglianese, Richard Craswell, Tino Cuellar, John Donahue, Bob Ellickson, Richard Fallon, Dan Farber, George Fisher,Ron Garet, Heather Gerken, Victor Goldberg, Bob Gordon, Jeff Gordon, Tom Grey, Joe Grundfest, Keith Hylton, SamIssacharoff, Dan Kahan, Pam Karlan, Michael Klarman, Dan Klerman, Susan Koniak, Gary Lawson, Sandy Levinson, JohnMcGinnis, Mike Levine, John Manning, Ed McCaffrey, John McGinnis, Frank Michelman, Martha Minow, Eric Posner,Mark Ramseyer, Bill Sage, Larry Sager, Fred Schauer, Margo Schlanger, Bob Scott, David Shapiro, Reva Siegel, Charles

i

Silver, Matt Spitzer, Rob Stavins, Jordan Steiker, Jeff Strnad, Bill Stuntz, Cass Sunstein, Eric Talley, Barry Weingast, andfrom other participants in the Law School Workshops at Boston University, Cardozo, Harvard, Michigan, Stanford, Yale,Texas, and the University of Southern California, as well as the Law and Economics Workshops at Columbia andHarvard, and the New Directions in Regulation Workshop at the Center for Business and Government at Harvard’sKennedy School of Government.

PREFERENCE-ESTIMATING STATUTORY DEFAULT RULES

By Einer Elhauge

I. INTRODUCTION

II. SHOULD STATUTORY DEFAULT RULES BE DESIGNED TO MAXIMIZE POLITICAL SATISFACTION?A. Political Satisfaction v. Judicial JudgmentB. Political Satisfaction v. Statutory Coherence, Stability or CertaintyC. Preference Satisfaction v. Fostering DeliberationD. At Least An Important Goal

III. ENACTING PREFERENCE-ESTIMATING STATUTORY DEFAULT RULES

A. Universal Default Rules Fitting Any Governmental PreferenceB. Tailored Default Rules Reflecting Legislative History

Objection 1. There Is No Legislative Intent. Objection 2. Legislative History Is Inaccurate. Objection 3. Using Legislative History Is Unconstitutional.

C. Updating for Changed Circumstances, Opting Out, and How Hermeneutic Inquiry DiffersIV. WHEN NO ONE OPTION IS MORE THAN LIKELY TO REFLECT ENACTABLE PREFERENCES

V. THE THEORY BEHIND A CURRENT PREFERENCES DEFAULT RULE

A. Current Preferences More Reliably AscertainableB. Current and Enacting Preferences Equally Ascertainable.

Objection 1:A Preference to Have Something to Run Against. Objection 2: Weaker Preferences Regarding Current Experience than Regarding theFuture Application of What One Enacted. Objection 3: Legislators Prefer Durable Legislation.

C. Enacting Preferences More Reliably Ascertainable.D. The Difference From Prior Rational Choice Models.E. Limits on a Current Preferences Default Rule

VI. TRACKING CURRENT LEGISLATIVE PREFERENCES IN PRACTICE

A. Subsequent Legislative Action Retaining or Relying on Otherwise NonbindingInterpretations.

B. Other Evidence of Changed Legislative PreferencesC. When to Overrule Statutory Precedent.D. Evidence That Legislatures Prefer a Current Preferences Default Rule

VII. DEFERENCE TO INTERPRETATIONS BY POLITICALLY ACCOUNTABLE AGENCIES

A. Chevron as a Current Preferences Default RuleB. Alternative Theories of ChevronC. Explaining Chevron’s LimitsD. Conflicts Between the Executive and LegislatureE. Conclusion on Agency Deference.

* Professor of Law, Harvard Law School. Email: elhauge@law.harvard.edu. I am grateful for funding fromHarvard Law School and the John M. Olin Center for Law, Economics and Business at Harvard Law School. I have alsobenefitted from comments on earlier versions of this paper by Scott Altman, Ian Ayres, Sam Bagenstos, Jack Balkin,Joseph Bankman, David Barron, Jack Beerman, Yochai Benkler, Victory Brudney, Thomas Campbell, Alex Capron, CaryCoglianese, Richard Craswell, Tino Cuellar, John Donahue, Bob Ellickson, Richard Fallon, Dan Farber, George Fisher,Ron Garet, Heather Gerken, Victor Goldberg, Bob Gordon, Jeff Gordon, Tom Grey, Joe Grundfest, Keith Hylton, SamIssacharoff, Dan Kahan, Pam Karlan, Michael Klarman, Dan Klerman, Susan Koniak, Gary Lawson, Sandy Levinson, JohnMcGinnis, Mike Levine, John Manning, Ed McCaffrey, John McGinnis, Frank Michelman, Martha Minow, Eric Posner,Mark Ramseyer, Bill Sage, Larry Sager, Fred Schauer, Margo Schlanger, Bob Scott, David Shapiro, Reva Siegel, CharlesSilver, Matt Spitzer, Rob Stavins, Jordan Steiker, Jeff Strnad, Bill Stuntz, Cass Sunstein, Eric Talley, Barry Weingast, andfrom other participants in the Law School Workshops at Boston University, Cardozo, Harvard, Michigan, Stanford, Yale,Texas, and the University of Southern California, as well as the Law and Economics Workshops at Columbia andHarvard, and the New Directions in Regulation Workshop at the Center for Business and Government at Harvard’sKennedy School of Government.

1

VIII. CONCLUSION

PREFERENCE-ESTIMATING STATUTORY DEFAULT RULES

By Einer Elhauge*

© 2002 Einer Elhauge. All Rights Reserved.

I. INTRODUCTION

Statutory interpretation involves two crucial issues. (1) How should courts divine the meaning ofstatutes? (2) How should courts decide what to do when they cannot divine a statute’s meaning? The firstissue raises fascinating issues of hermeneutics, and has both dominated the literature and been dominatedby debate about the extent to which interpretive theory can or should constrain judges to act as honestagents for the legislature. The second issue has suffered from relative neglect, with its typical answer –more assumed than justified – being that in resolving statutory indeterminancy, courts must exercise judicialjudgment to choose the substantively best results or canons. To the extent the second issue is engaged, thetypical argument is more about what the substantively best results or canons are , and whether they shouldbe framed as case-by-case standards or categorical rules, rather than about the premise that judicialjudgment is how to resolve the question.

I here take a different tack. I will skip the first issue and instead begin wherever hermeneuticsleaves off by focusing exclusively on how courts should resolve admitted statutory ambiguities. Further,regardless of the extent to which hermeneutic theory constrains judges to act as honest agents, I will argue

1 See, e.g., Cass R. Sunstein, Interpreting Statutes in the Regulatory State, 103 HARV. L. REV. 405, 413, 415-41(1989) [hereinafter Sunstein, Interpreting Statutes] (“traditional sources offer incomplete guidance and . . . theirincompleteness reveals the inevitable failure of the agency conception of the judicial role.”)

2 I will throughout use the term “meaning” to refer to whatever meaning or range of possible meanings resultsfrom the reader’s favored interpretive methodology, which to some turns on the statutory text and linguistic conventionsbut to others turns on statutory “purpose,” or legislative “intent” or “will.” I mean the term “meaning” to be agnosticon that and all other hermeneutic issues of proper interpretation. I will also be abstracting from subtle hermeneuticdistinctions about whether a particular form of statutory indeterminancy is best called a ‘gap,’ an ‘ambiguity,’‘vagueness,’ a ‘conflict,’ an ‘omission,’ ‘unclarity’ or something else. I will instead use all those terms interchangeablyto mean that whatever hermeneutic approach the reader favors has failed to resolve the question of meaning. Forexample, if one’s favored interpretive method provides that a true statutory “gap” indicates the statute meant to excludethat case, then that person would see no indeterminancy creating the occasion for a default rule. But others might thinkthat at least some gaps (like a failure to specify a statute of limitations) do create indeterminacy that judges must resolve,so I will sometimes refer to gaps as a possible basis for a default rule.

2

that in resolving any statutory ambiguities left after interpretive inquiry, courts should not exercise judicialjudgment but rather should act as honest agents for the political branches. This may seem impossible since,by definition, these are the set of cases where the legislative instructions are unclear. And so many havethought.1 But courts can still act as honest agents by adopting a set of statutory default rules designed tominimize political dissatisfaction with statutory results. In fact, I will argue that they largely have done justthat, and that this explains various seeming anomalies in statutory interpretation. But, and quitecounterintuitively, I will show that the statutory default rules that minimize political dissatisfaction often donot track the most likely meaning or even preferences of the enacting legislature.

Given my focus, I intentionally abstract not only from the debate about the extent to whichhermeneutic inquiry constrains interpretations of statutory meaning, but also from interesting and importantdisputes about how best to conduct any hermeneutic inquiry: whether to look just to text or to legislativehistory, dictionaries, and current practice; whether to focus on textual meaning or legislative intent orinferred statutory purpose; and whether and how to “translate” or otherwise consider the effect on meaningof unforeseen applications or changing times, circumstances, or context. I rely rather on the simple premisethat at some point these interpretive efforts must give out. No matter how much faith one has in the latesthermeneutic theory, sometimes we cannot determine statutory meaning through interpretative efforts.Interpretive theory may generally leave us with a range of possible meanings, but often cannot resolve whichmeaning to choose.2 The need for a doctrine to resolve this indeterminacy only increases if one issufficiently agnostic about hermeneutic theory to believe judges and scholars can reasonably subscribe todifferent theories, for in many cases reasonable hermeneutic theories point in the opposite direction. Buteven a partisan for a particular hermeneutic theory should confess that reasonable persons applying thatsame theory would often reach different conclusions in the same case -- that is, the theory has manyambiguous applications that cannot be resolved by the theory itself.

In short, dear reader, in whatever set of cases you are willing to concede present statutoryindeterminacy, what statutory doctrine should provide the default rules that resolve those indeterminacies?The importance of this question surely varies with how determinate our methods for divining statutorymeaning are. To those who find those methods wholly indeterminate, the choice of default rules actually

3 Outside the hermeneutic debate, rational choice theorists have in their models tended to assume that in makingstatutory decisions judges simply maximize their own preferences, without making any value judgment about whetherjudicial preference-satisfaction is desirable or not. See WILLIAM ESKRIDGE, DYNAMIC STATUTORY INTERPRETATION 167-70(Harvard Univ. Press 1994) [hereinafter ESKRIDGE, DYNAMIC]; William N. Eskridge & John Ferejohn, Making the DealStick, 8 J. L. ECON. & ORG. 165, ___ (1992); William N. Eskridge, Jr. & John Ferejohn, The Article I, Section 7 Game, 80GEO. L. J. 523, 548-51 (1992) [hereinafter Eskridge & Ferejohn, Game]; Spiller, supra note , 12 INT. REV. L. & ECON. at 187-88; Gely & Spiller, A Rational Choice Theory of Supreme Court Statutory Decisions with Applications to the State Farmand Grove Ci ty Cases , 6 J.L. ECON. & ORG. 263, 267 (1990); Bill Eskridge, Overriding Supreme Court StatutoryInterpretation Decisions, 101 YALE L.J.. 331, 378-85 (1991) [hereinafter “Eskridge, Overriding”]; Schwartz, Spiller andUrbiztondo, A Positive Theory of Legislative Intent, 57 LAW & CONTEMP. PROBS. 51, 56-57, 72 (1994); see also LindaCohen & Matthew Spitzer, Solving the Chevron Puzzle, 57 LAW & CONTEMP. PROBS. 65, 71-76 (Spring 1994) (assumingthose judicial views incorporate preferences about both policy and legal process). Since these models begin with thisdifferent premise, they not surprisingly reach different conclusions from mine.

4 I hope to persuade such scholars that, even if they are correct in their hermeneutic skepticism, a properlydefined set of statutory default rules may make the condition that they regard as lamentable less inescapable than theyhave presupposed.

3

resolves more cases than the choice of interpretive theory – at the extreme all cases. But even if one hasa high degree of faith in interpretive theory, the choice of statutory default rules will resolve many interestingcases in whatever residual remains unresolved by interpretive theory. In Supreme Court cases, the residualwill be particularly large, and thus default rules particularly important, because cases that do not involvestatutory ambiguity are less likely to split the lower courts.

The hermeneutic debate that has dominated the statutory interpretation literature has generallyassumed that what must resolve any indeterminacy left by hermeneutic theory is judicial judgment. But farfrom producing consensus, this common assumption has only sharpened divisions because of differingreactions to such judicial judgment. Some fear it, some like it, and others simply think it inevitable. Thosewho fear it regard judicial decisionmaking as illegitimate or undesirable because it is politicallyunaccountable. This first camp tends to focus on hermeneutic efforts to constrain judicial judgment bydefending and improving methods of divining statutory meaning (and judicial compliance with them) thatproduce more determinate results. Their assumption (and fear) is that unless constrained by hermeneutictheory, statutory gaps will be filled in by judicial judgment. Those who like judicial judgment instead tendto emphasize the weaknesses of hermeneutic methods of divining statutory meaning. Rather than seek toconstrain judicial judgment, they celebrate it. Some in this second camp think the judicial process betterprotects certain fundamental values or traditions, in part because of its system of precedent and commonlaw development. Others in the second camp stress that the judicial process is more nimble, aware ofchanged circumstances, and focused on fact-specific applications. But they share the basic premise (thoughhere hope) of the first camp: that statutory issues unresolved by hermeneutic theory will be left to the judicialprocess. A third camp simply thinks that because hermeneutics are not very constraining (or often are not),judges have no choice but to exercise judicial judgment or maximize judicial preferences when decidingstatutory issues. Those in this camp may be neutral on this fact,3 or regard it as lamentable,4 but theyassume it is inevitable.

All these diverse camps thus share the premise – often unspoken and rarely analyzed – that judicial

5 Although hypothetical consent and penalty default rules have been popularized and systematized by recentlaw and economics scholarship, see, e.g., Ian Ayres & Robert Gertner, Filling Gaps in Incomplete Contracts: AnEconomic Theory of Default Rules , 99 YALE L.J. 87 (1989), in fact these notions go back at least to Lon Fuller. See FULLER

& BRAUCHER, BASIC CONTRACT LAW 557-58 (1964) (noting contract law is designed either to reflect reasonableexpectations or to induce parties to clarify them).

6See Ayres & Gertner, supra note , at 129-30 (article laying out theory of contractual penalty defaults that, inits conclusion, noted its analysis might be extended to statutes too); Elizabeth Garrett, Legal Scholarship in the Age ofLegislation, 34 TULSA L.J. 1, 4, 8-9 (1999) (noting without endorsing this possibility); Adrian Vermeule, InterpretiveChoice, 75 N.Y.U. L. REV. 74, 85 (2000) [hereinafter Vermeule, Interpretive Choice] (noting without endorsing). Perhapsthe most extensive prior treatment is in Cass Sunstein, Must Formalism Be Defended Empirically?, 66 U. CHIC. L. REV.636, 645-50 (1999), which outlines this possibility and agrees it provides “at least a reasonable starting point,” id. at 649. But Sunstein ultimately does not endorse this approach, id., and indeed at least partially rejects it to the extent he alsoargues for the traditional position that interpretation should also be affected by judicial judgment about public policyvalues. Id. at 650; Sunstein, Interpreting Statutes, supra note , at 413, 466-67, 476-89.

7 As I use the term, “enactable political preferences” are the political preferences of the polity that are sharedamong sufficient elected officials (taking into account any requirements for the concurrence of different political bodieslike a House and Senate or President) that they could and would be enacted into law if the issue were on the legislativeagenda. Enactable political preferences thus does not refer to polling data or other indications of the polity’s generalpolitical preferences that are not manifested in their choice of elected officials, nor to strategic private aims that

4

judgment is what does and must govern after hermeneutics gives out. To be sure, many articles and caseshave begun to recognize the existence and necessity of statutory default rules (often by another name) toresolve statutory ambiguities. But they tend to base those rules on substantive judgments that they believehave or should be adopted by the judicial process. These articles and cases thus do not really deviate fromthe premise that indeterminate meaning must be resolved by judicial judgment, but rather favor a more rule-like implementation of judicial judgment. None of the dominant camps has focused seriously on thepossibility that ambiguities about statutory meaning could instead be resolved by default rules that constrainjudge to maximize political preference satisfaction.

This common premise of statutory interpretation is interesting because the situation is dramaticallydifferent for judicial interpretations of another kind of legal text: contracts and corporate charters. In casesand scholarship concerning contracts and corporate law, it has long been openly acknowledged thatinterpretive methods used to divine the meaning of contracts and corporate charters often run out. Andthis caselaw and literature generally assumes that what should fill these ambiguities or gaps in meaning arenot exercises of judicial judgment, but rather whatever default rules most accurately reflect or elicit thepreferences of parties agreeing to such contracts or charters.5 Given this, a small but growing set ofscholarship has suggested the contractual default rule approach might be applied to statutory interpretation,though often without endorsement, and certainly without yet working out any systematic defense of thisproposition or account of what set of statutory default rules would maximize legislative preferencesatisfaction.6

I aim to provide that systemic defense and account. More precisely, I aim to explain how adoctrine of statutory default rules should (and currently generally does) constrain judges to maximize theextent to which statutory results accurately reflect enactable political preferences, as measured by whateverpolitical process is accepted in that society for enacting statutes.7 One point the analysis will demonstrate

legislators may harbor but could not actually enact into law.8 See generally FALLON, MELTZER & SHAPIRO, HART & WECHSLER’S THE FEDERAL COURTS AND THE FEDERAL

SYSTEM 755-57 (4th ed. 1996) (describing the range of views on how to classify federal common law versus statutoryinterpretation). Sometimes, though, the political preference will be to delegate the issue to be resolved by thesubstantive judgment of courts or agencies. Id. at 754; infra at Part II.A; Elhauge, Preference-Eliciting Statutory DefaultRules, at II (on file with author) [hereinafter “Elhauge, Preference-Eliciting”].

9 See, e.g., Stephen F. Ross, Where Have You Gone Karl Llewellyn? Should Congress Turn Its Lonely EyesTo You?, 45 VAND. L. REV. 561, 566-72 (1992) (suggesting that Congress adopt such interpretive rules by statute withoutspecifying the content); [new Harvard Law Review piece on federal rules of statutory interpretation] (same).

10 Generally it will be in the self-interest of legislatures to have statutory default rules that maximize thesatisfaction of enactable political preferences. But there will also be limits because legislators might try to maximizecareerist self-interest over the satisfaction of enactable preferences, or try to change default rules to favor its interestsor views over those of future legislatures. See infra Part V; Elhauge, Preference-Eliciting, supra note , at III.D. Thoselimits might even be deemed constitutionally mandatory based on an interpretation of the legislative and judicial powersunder the relevant Constitution, but in any event they set a limit on my recommendations.

11 I will not here be making or addressing any claim that my recommended statutory default rules are logicallynecessary or otherwise inevitable as a matter of hermeneutics itself. Further, although I extend my normative theory toother nations, I have not yet undertaken the systematic investigation necessary to establish whether statutory doctrineelsewhere also fits my descriptive theory.

5

is that, while the contract and corporate law approach of resolving textual gaps or ambiguities with defaultrules that maximize the preferences of the contracting parties is also fruitful for statutory texts, one mustmake many modifications because statutes raise unique issues.

In developing this account, I also aim to challenge the widespread premise that hermeneuticinsufficiency means some substantive judicial judgment must resolve statutory indeterminacy. Or, to phrasethe proposition in an alternative way that some may prefer, wherever you believe statutory interpretationleaves matters to judicial judgment, my argument is that courts should exercise (and largely have exercised)their judicial judgment to adopt default rules that maximize political satisfaction. This is true whether onecategorizes the judicial judgment as an aspect of statutory interpretation or as an exercise of an interstitialcommon law power created by the relevant statute.8 Further, if one instead focuses on the possibility thatlegislatures might themselves enact statutory default rules binding on the courts in cases of statutoryambiguity,9 my analysis provides a recommended content for (and limits on10) the rules the legislature shouldadopt. In fact, as we will see, although the interpretive rules provided by statute in the United States Codeand many state codes are inconsequential, many other states have enacted more substantive codes ofconstruction that provide valuable evidence on the default rules most likely to maximize political satisfaction.

Throughout I will be making both the normative claim that my theory is desirable, and thedescriptive claim that it largely fits the U.S. case law.11 I do not by this descriptive claim mean that mytheory matches what judges subjectively think they are doing or say they are doing in their opinions.Rather, I mean that my theory fits and predicts the doctrine. Indeed, designing default rules to minimizepolitical dissatisfaction explains and justifies many judicial practices, doctrinal distinctions, and canons ofconstruction far better than existing interpretive or substantive theories. It also helps resolve what mightotherwise appear to be little more than open-ended conflicts among statutory canons and cases. This

6

improved understanding of the reasons underlying statutory constructions, and the justifiable grounds fortheir seemingly inconsistent application, also renders them more determinate, and thus more constrainingof judges.

The normative and descriptive claims stand separately, but draw additional strength from theircombination. They stand separately in the sense that, if one disagrees with the normative thesis, my claimwould remain that this theory best explains the actual contours of current statutory interpretation. Likewise,if one disagrees with my descriptive thesis, my point would remain that courts or legislatures shouldexercise their power over statutory interpretation to adopt statutory default rules that maximize politicalsatisfaction, and that legislatures should require them to do so if they prove unwilling.

But my argument also invokes that combination of descriptive and normative claims that is thepeculiar province of law professors. Because one of the tasks of legal scholarship is to explain legaldoctrine in a way that can provide guidance to future courts, it has always been important to establish thata sufficient fit exists between extant doctrine and the proffered justification. For example, suppose oneconcluded that my descriptive claim was not as accurate as the contrary claim that judges simply interpretstatutes in whatever way furthers their personal ideological preferences. Even if this offered a viablepositive account acceptable in political science and rational choice theory, such a normatively corrosiveposition cannot offer an attractive legal theory for guiding future courts. A valid legal theory must insteadhave some normative justification to merit adoption. Conversely, a perfectly valid normative argument thathas no connection to existing doctrine may offer a useful blueprint for legislative reform, but has no claimto being a theory of legal doctrine. Without any grounding in the authority of existing doctrine, such anormative theory cannot offer guidance to an agency, trial court, or intermediate appellate court. Its utilityis even limited before a jurisdiction’s high court, partly because of the presumption in favor of stare decisis,but even more so because that court must rule case-by-case in reviewing lower courts that will generallybe following existing doctrine. This tends to limit high court decisions to altering the margins of existingdoctrine and makes it difficult to accomplish a wholesale shift to an entirely different normative foundation.Thus, while legal theory often includes pure positive or normative theory, what distinguishes it from thoseendeavors is that it also focuses on determining which of the possible normative justifications is mostconsistent with the descriptive landscape. The best legal theory might thus be neither the most descriptivelyaccurate nor the most normative attractive, but rather can be the theory that provides the best combinedfit of descriptive explanation and normative justification. As well as making the separate claims, I makeprecisely such a combined claim on behalf of my legal theory of statutory interpretation.

I begin by justifying, in Part II, the claim that statutory default rules should maximize the extent towhich statutory results accurately reflect enactable political preferences. I defend this claim againstarguments that courts should instead advance other conceptions of the good, statutory coherence, orlegislative deliberation. I then develop my counterintuitive conclusion that, where statutory meaning issubstantially unclear, an approach of maximizing political satisfaction often dictates adopting canons ofstatutory construction that do not reflect the enactors’ most likely meaning – or sometimes even theirpreferences. I develop this point in three stages.

In the first stage, considered in Parts III-IV, I assume there is no reason to think that the political

12 By the “current” government, I mean the one that exists when the court interprets the statute. The “enacting”government is the one that existed when the statutory text was enacted. I will often use the term “government” (or“enactors”) rather than the term “legislature” to emphasize that it includes all those who affect which political preferencescould be enacted even if (like a President or governor under U.S. law) they are not technically part of the legislature.

13 See Elhauge, Preference-Eliciting, supra note __.

7

preferences of the enacting and current government differ,12 nor that the political process would respondto any statutory interpretation by enacting a different statute. In this case of static non-responsivepreferences, the statutory default rules should be chosen to maximize the satisfaction of the enactinggovernment’s preferences. These enacting preference-estimating default rules come closest to parallelingthe meaning that most likely would have been attached to indeterminate text by the enacting government.But the inquiries differ, mainly because the sources of information for making probabilistic estimates ofpolitical preferences are broader than those for divining the meaning the enactors likely attached to aparticular text. Further, given sufficient uncertainty about which preferences are enactable, minimizing thedissatisfaction of those preferences will often turn out to dictate adopting moderate interpretations evenwhen more extreme interpretations are more likely to be enactable.

In the second stage, addressed in Parts V-VII, I add the complication that the political preferencesof the enacting and current governments may differ. Contrary to ordinary supposition, in such cases, thedefault rule that overall best maximizes the political preferences of the enacting government will turn outto be tracking the preferences of the current government where they can be reliably ascertained fromofficial action. This argument for a current preferences default rule may seem counterintuitive -- would notthe enacting government want its own political preferences followed? The answer would be ‘yes’ if weasked the hermeneutic question: that is, what meaning would that enacting government most likely want forthat particular statute? But the answer is ‘no’ if we instead ask the default rule question: that is, whatgeneral rule for resolving indeterminacies about statutory meaning -- including those in older statutes beinginterpreted and applied during the time that the enacting government holds office – would most maximizethe political satisfaction of the enacting government?

In the third stage, I add the complication that statutory interpretations (or the prospect of them) canthemselves often elicit a legislative reaction. In a limited set of cases, it turns out political satisfaction canbe maximized by choosing what I will call a preference-eliciting default rule, which intentionally differs fromlikely political preferences in order to elicit a political response that will make it clearer just what thegovernment desires. In those cases, a preference-eliciting default rule can create statutory results thatreflect enactable preferences more accurately (and deliberatively) than any judicial estimate possibly could.This proposition is established in a companion piece, which analyzes the necessary conditions and explainshow many canons of statutory construction can best be explained on these grounds.13

In this article, however, I will instead focus on how to arrive at a statutory default rule for thecommon case where the proper conditions do not exist for choosing a statutory interpretation that elicitsa legislative reaction. Beginning with the simplest case of static preferences permits us to focus on thedistinction between the normal hermeneutic question (what indications do we have about what the enactorsmeant or intended?) and the default rule question (what rule would maximize the enactors’ political

8

preferences?). The latter, Part III demonstrates, provides a more solid foundation for traditional doctrineslike the canon against absurd interpretations, and modern proposals to update statutory interpretationswhen circumstances change, each of which sit uneasily with hermeneutic theories of meaning. That is, evenif one is not convinced by hermeneutic theories that courts can disregard a statutory meaning that seemsto have become absurd to the judge given unanticipated or changed circumstances, a default rule that askswhat will maximize the satisfaction of the enacting government’s political preferences would certainly avoidabsurd results and vary with changed circumstances.

An enactable preferences default rule approach also provides entirely different – and morepersuasive – grounds for relying on legislative history. Even if critics are correct that legislative history doesnot establish statutory meaning or legislative intent, it helps make at least a probabilistic estimate of thepolitical preferences that were influential enough to be enactable. As long as those probabilistic estimatesare better than random, this will suffice to minimize political dissatisfaction even if, as many argue, legislativeintent is a incoherent concept. Indeed, even if legislative history is often inaccurate, a default rule of relyingon it will elicit corrections on the legislative record. Further, default rule analysis suggests that where inquiryinto legislative history is necessary to resolve a statutory ambiguity, the inquiry should extend beyond thelegislative history of the particular statute. It should include any other debates or information bearing onthe relevant political preferences, including the enactment of other statutes and general historical informationabout the political forces of that period.

Part IV then explores an ambiguity under default rule analysis: what should courts do when no oneoption more likely than not matches enactable preferences? This ambiguity has never been resolved by thecorporate and contract law scholarship on default rules. The analysis here shows that, where no one optionis more than 50% likely to be enactable, courts should not just choose the plurality option (i.e., the optionwith the highest likelihood of matching enactable preferences). Instead, somewhat surprisingly, it minimizesexpected political dissatisfaction to choose a moderate interpretive option even though extreme options aremore likely to match enactable preferences. This is not a claim that the moderate option would most likelyhave been needed to pick up the median legislator. Even if one predicts the median legislator would bemore likely to pick an extreme option, a more moderate option should be chosen when the extreme optionis less than 50% likely to reflect enactable preferences. Likewise it would minimize the expecteddissatisfaction of contracting parties or corporate participants to adopt a similar “moderate” default rule forsimilar cases of contract or corporate law interpretation.

Moving to the second stage allows us to consider the implications of changed political preferences.From the perspective of the enacting government, as detailed in Part V, a current preferences default rulemakes future statutory results correspond less well to its preferences, but makes present statutoryinterpretation of old statutes correspond better. Which will be more important to the government: satisfyingits political preferences regarding the future or the present? As a general matter, political preferences fora given statutory result are likely to be stronger in the present because those who hold those preferences(and elect the government) are those who experience that result. Only some of them will still be aroundwhen the future statutory result comes (in some cases none), and even if they are still around, many willhave different political views or be affected differently by the statutory result because their life-situation haschanged. They are thus likely to have weaker preferences about whether their favored statutory result

14 See infra at V.

9

obtains in the future than in the present. Additionally, the set of old statutes being applied in the presentwhen the enacting government sits will cover the full range of statutory results that affect their electorate’sexperience, whereas the enacted statutes being applied in the future will be smaller in number and coversome more narrow range of life experiences. Thus, even for the enacting government, a general default rulethat accurately tracks current preferences (rather than the preferences of the government that enacted eachstatute) will maximize its political satisfaction.

The default rule that best maximizes the preference satisfaction of the enacting government will turnnot just on the relative strength of its preferences about the present and future, but on the relative accuracywith which those preferences can be ascertained. If the enactable preferences of the enacting governmentare unknown but can reliably be estimated for the current government, then an enacting government wouldsurely prefer a default rule that follows current preferences. For in such circumstances, such a default rulewould give the enacting government more political satisfaction with statutory interpretations made duringits time in office without sacrificing any influence over future interpretations. Likewise, when enactinggovernmental preferences are not entirely unknown but are less clear than current governmentalpreferences, that would also cut toward the enacting government preferring a current preferences defaultrule. But when enacting governmental preferences are clearer than current governmental preferences, thenthere will be conflicting effects. In such circumstances, a default rule of following current governmentalpreferences will give the enacting government more influence during a time its preferences are likely stronger(its time in office) but not during the time when they can be ascertained more accurately (the futureinterpretation of statutes it enacted). The enacting government would thus want a default rule that followscurrent enactable preferences only when they can be ascertained relatively accurately and reliably. Indeed,the current government would share this conclusion since it faces the same calculus given that the questionis which general default rule to pick. This reasoning provides a much more solid (and limited) ground supporting proposals for updatingstatutes to take into account changed political preferences, which some prominent scholars have made ondifferent grounds.14 We need not rely on a controversial conclusion that current political preferences arebetter or should take precedence over enacting preferences. Rather, we can rely on the ground that bothcurrent and enacting governmental preferences can often better be satisfied by a general default rule thatmaximizes the preference satisfaction of each during their time of rule. But this justification is limited tocases where statutory meaning is actually unclear, and the current political preferences tracked are thosemost likely to be enactable. The approach here thus rejects the prior proposals, which would allow judgesnot only to modify statutory meaning itself, but to exercise substantive judicial judgment in choosing amongthe unenactable positions held by current public opinion or values or in determining what the public wouldenact without political obstacles that, although constitutional, the judge deems undesirable.

Instead, the argument here justifies dynamic interpretation that reflects changing political preferencesonly when it is both limited to the resolution of statutory ambiguities and tracks the preferences that are themost likely to be enactable under the actual political system. Concerns about the reliability of ascertainingcurrent enactable preferences and about making sure that interpretations are stable enough to induce some

10

behavioral reliance further suggest that judges should follow only those current preferences that have beenmemorialized in some official action. Not surprisingly, when enacting preferences are obscure or ancient,courts are more relaxed about how certain their estimate of current enactable preferences must be sincethe enacting governmental preferences will be weak.

This default rule analysis, as shown in Part VI, provides a sounder explanation for (and betterexplains the limitations on) judicial treatment of subsequent legislative action and inaction, decisionsoverruling statutory precedent, and decisions like Bob Jones that resolve statutory ambiguities with acurrent legislative policy where it is sufficiently clear even though it conflicts with likely enacting legislativepreferences. It further provides an improved way of understanding Chevron deference to agencyinterpretation, and the various nuanced doctrines limiting such deference. Part VII analyzes those doctrinalnuances in some depth, including providing a fresh justification for the Mead doctrine and other denials ofChevron deference that have so far seemed perplexing under traditional Chevron theories.

II. SHOULD STATUTORY DEFAULT RULES

BE DESIGNED TO MAXIMIZE POLITICAL SATISFACTION?

A. Political Satisfaction v. Judicial JudgmentThe guiding principle throughout my analysis will be that statutory default rules should be devised

to maximize the satisfaction of political preferences, which necessarily entails that they will to some extentconstrain judges from imposing their own views or judgments. To say anything is constraining of judgesis, of course, to immediately invite postmodern deconstruction and dismissal of the entire enterprise asbased on a false assumption that these default rules can (any more than any other rules) be so precise asto prevent a judge from interpreting them in some other way. Those who share such concerns should feelfree to think of these as default axioms rather than rules to take account of the fact that they are unlikelyto be 100% determinate in individual cases. Indeed, many of the default doctrines are not that rule-like,but rather offer guidance whose content will be determined by context. But an axiom or rule need not be100% determinate to have a statistically significant effect on likely judicial outcomes. That is all the effectI will need to make this inquiry worthwhile.

True, a clever and dedicated subverter of default rules or axioms (as with any other rule orprinciple) can no doubt manipulate them to reach whatever result they personally prefer in most cases. Soit is with any instruction a principal gives to any agent. But it is hard to see why the political process wouldwant to appoint the kind of judicial agents who misuse their interpretive powers to pursue their own politicalviews or judgments. To the contrary, the political process would want to commit itself to a strategy ofrefusing to appoint -- and, where feasible, reversing the statutory decisions of -- judges who just tried tofurther their own political views or judgments, for commitment to such a strategy would maximize theexpected influence of those in the political process. Some mistaken judicial appointments will no doubt slipthrough the screening process, but over the entire range of judges and cases, identifying the correct defaultrules/axioms should increase the extent to which statutory results accurately reflect enactable politicalpreferences. If the judiciary is sufficiently wayward, the legislature can also try to reestablish the supremacyof democratic choice by enacting statutes that specify statutory default rules that maximize political

15 See supra note _.16 For theories that interpretation should follow the underlying “moral reality,” see Michael S. More, A Natural

Law Theory of Interpretation, 58 S. CAL. L. REV. 277, 353-58, 383-86 (1985); Heidi M. Hurd, Challenging Authority, 100YALE L.J. 1611, 1620, 1667-77 (1991); Heidi M. Hurd, Sovereignty in Silence, 99 YALE L.J. 945, 995-97, 1028 (1990). Analternative version of this objection might say that interpretation must vindicate people’s “rights.” But absent someclaim of privileged insight into morality reality, to say there is a “right” is to presume we have some defined constitutionalconstraint or statutory meaning, and this paper assumes we are operating in the realm where we have neither.

11

satisfaction.15

Alternatively, the dedicated rule skeptic can think of the question under study as simply aspirational.Whether or not constraining statutory rules or doctrines can be designed, what kind of judicial practicewould citizens of a democracy want judges to follow when filling in statutory defaults? And what kind ofpractice would honest interpretive agents want to follow? The general answer seems plain. Democraticcitizens would want a judicial practice of applying those default rules/axioms that maximize the extent towhich statutory results accurately reflected their political preferences.

One might object to this even as an aspiration. Why shouldn’t judges do the right thing no matterwhat the legislative polity thinks?16 I begin with two institutional responses. The first stresses that it is astatute we are expounding. The whole reason for having a political process for enacting statutes is todetermine what the right thing is by assuring that, within constitutional bounds, political preferences arereflected in statutory results. If we preferred a judicial process, the enactment of statutes could beconstitutionally delegated to judges instead. Absent that or some constitutional constraint, the statutoryresults that resolve statutory ambiguities should reflect enactable political preferences, as measured by theaccepted process for enacting statutes in that society. Nor does the necessary task of statutoryinterpretation provide any warrant for judges to impose their views of how democracy might be “perfected”by interpreting statutes to reach results that would be enactable only if one lifted certain obstacles that, whileconstitutional, the judge finds objectionable. If a society has imposed legislative obstacles (like theconcurrence of separate legislative houses and an executive) to alter the extent to which politicalpreferences can be translated into statutory enactments, then considering those legislative obstacles shouldalso alter the extent to which political preferences can be translated into statutory interpretations. If thereis some unconstitutional defect in the process used to enact statutes, then that process must be invalidatedfor statutory meaning and ambiguity alike. If the alleged defect is constitutionally valid, then it is up to thesociety to determine whether it should reform that process to remove the defect. But as long as theenactable preferences measured by the legislative process legitimate compelled obedience to the statuteitself, those same set of preferences should govern the statutory interpretations that command the sameobedience.

My second institutional response stresses that statutory interpretation is a collective act, not anindividual assessment. The project is to advise interpreters (normally judges) how best to makeinterpretations that will be collectively binding on us all. The judges who disagree with any scholar’sproffered normative assessment about what “the right thing” is have no reason to adopt that scholar’sadvice about the best statutory default rules. The other judges might adopt the advice, but then the questionbecomes why these judges should be able to impose that contested normative assessment on citizens who

17 Perhaps there are some normative assessments that are uncontested, but if they exist they just represent aspecial set of interpretations that maximize political satisfaction, so adopting them is entirely consistent with myapproach. The more interesting and typical cases involve contested normative assessments.

18 See Tribe, The Puzzling Persistence of Process-Based Constitutional Theories, 89 YALE L.J. 1063, 1072-76(1980) (establishing proposition for claims of that a minority is under-represented); Elhauge, Does Interest Group TheoryJustify More Intrusive Judicial Review?, 101 YALE L.J. 31, 48-66 (1991) [hereinafter Elhauge, Interest Group Theory](establishing proposition for claims that majority is under-represented).

12

disagree with it?17 And why should statutory interpretation turn on the happenstance of which normativeviews happen to be held by the judge one draws? Normative theory provides a powerful tool by whicheach of us can individually decide which statutory results are undesirable. But it cannot usefully providean independent basis to make collective decisions (through judges) about how to interpret the results ofa collective process that itself is supposed to decide what is normatively desirable. Nor are judges’normative views about substance and process really separable. No one can determine whether a minorityor majority group has too little political influence because of process defects without determining how muchinfluence that group normatively ought to have.18 Thus, any judicial conclusion to favor a group in statutoryinterpretation because it is under-represented by the actual legislative process amounts to judicial impositionof the substantive normative views that led to that conclusion.

The root problem underlying these two institutional responses is that we all harbor differentconceptions of what the right thing is. Each person’s ideal system would adopt their conception, buteveryone cannot be a dictator on conceptions of the social good. The best we can do is join a democracythat accurately and equally weighs our different conceptions. True, a democracy will often want to putlimits on what it can do, and even more so on what its political agents can do. But then we are in the realmof constitutional law rather than statutory interpretation. Throughout this paper, I will assume judges areoperating in whatever area is left open by constitutional constraints. In choosing among those statutoryresults that are constitutional, the legislative polity will want those results that minimize its expected politicaldissatisfaction.

Why should any of us accept the statutory results of such a democracy even when they violate ourconception of justice and the social good? Because, in exchange, others accept statutory results we wouldregard as good but they would regard as bad. We might justify such a tradeoff on consequentialist grounds:if everyone accepts this social compact, the expected good consequences outweigh the bad for each ofus. After all, the alternative to accepting the statutory results of a democracy that sometimes violatesindividual conceptions of the good is to accept the statutory results of a non-democracy that violatesindividual conceptions of the good more often. Unless, that is, the society does not accept statutory resultsas authoritative at all, and that way lies not just chaos but the frustration of our collective ability to pursueanyone’s conception of the good.

Non-consequentialist arguments about democracy support the same conclusion. Democratictheories that center on respecting the majority’s will would obviously conflict with judges adopting statutoryinterpretations that correspond with their personal views or judgments rather than majority preferences.But the same follows even if one rejects majority will in favor of democratic theories that instead stress the

19 See N. BOBBIO, DEMOCRACY AND DICTATORSHIP 137 (P. Kennealy trans. 1989); HANS KELSEN, GENERAL THEORY

OF LAW AND STATE 284-86 (A. Wedberg trans. 1961).20 Even if one regards any particular statutory result as only a temporary “institutional settlement” in an ongoing

debate about conceptions of justice and the common good, see Hart & Sacks, [cite], the question remains whichtemporary institutional settlements should be reached, and temporary settlements that track political preferences willminimize the extent to which the polity finds its conceptions of the good violated at each point in time.

21 I do not by this discussion mean to be drawing any sharp distinction between political preferences andconceptions of justice. Both doubtless play a role in the enactment of statutes. I rather mean that whatever one callsthe reasons that legitimate compelled obedience to clear statutes (be they conceptions of justice, preferences, opinions,views, or judgment), those same reasons should govern interpretative default rules in cases of statutory unclarity.

22 See FALLON, MELTZER & SHAPIRO, supra note , at 754; Elhauge, Preference-Eliciting, supra note , at II.23 See Business Electronics v. Sharp, 485 U.S. 717, 732 (1988); Easterbrook, Statutes’ Domains, 50 U. CHIC. L.

REV. 533, 544 (1983). 24 RONALD DWORKIN, LAW’S EMPIRE 327-54 (Harvard University Press 1986).

13

principle of self-governance.19 For being a participant in a self-governing democracy necessarily meansgiving up a claim to privileged insight into what advances the public interest. It means committing insteadto have the public interest determined through some political process of preference aggregation orprincipled deliberation that constitutes the collective means by which the polity governs itself.

What democratic government cannot mean is producing a government where everyone'sconception of the good is furthered, because that is impossible. All it can produce is the best that ispossible: a government that minimizes the expected dissatisfaction of the polity's differing conceptions ofthe good.20 But it can only do so if interpretive agents apply default rules that advance this goal rather thanmisuse their interpretative powers to further their own conceptions of the good. An interpreter that insistsher conception of the good must take precedence breaches the fundamental social contract that leaves suchmatters to be resolved democratically.21

The default rule approach here thus rejects – and should be contrasted with -- the alternativedefault rule (implicit in many theories) that leaves all statutory indeterminacy to be resolved by judicialjudgment. Courts should choose not what they regard as wise policy, but should instead choose defaultrules that on average minimize political dissatisfaction.

To be sure, sometimes it will be clear that what the legislature meant or preferred was to delegatethe matter for ongoing judicial resolution.22 The antitrust statutes are, for example, commonly understoodto involve such a delegation to courts.23 But in such a case, exercising substantive judicial judgment doessatisfy legislative preferences, and thus differs from the exercise of such judgment even when no suchdelegation was preferred.

B. Political Satisfaction v. Statutory Coherence, Stability or CertaintyThe objection might continue that, even if judges should not impose their own views, it does not

follow that they should maximize the satisfaction of political preferences. Ronald Dworkin’s influentialwork has argued that courts should instead maximize the extent to which legislative results can be justifiedby a coherent and principled set of general political convictions.24 But on Dworkin’s premise thatlegislators share his hierarchy of preferences -- which make general political convictions more important

25 Id. at 329-337.26 Id. at 330-31. 27 Id. 28 DWORKIN, supra note , at 332-33. 29 Id.30 See, e.g., CAL. CODE CIV. PROC . §1859 (“when a general and particular provision are inconsistent, the latter

is paramount to the former. So a particular intent will control a general one that is inconsistent with it.”); MONTANA CODE

§1-2-102 (same); OREGON REV. STAT . § 174.020 (same); COL. REV. STAT . §§ 2-4-205 (when irreconcilable a specificprovision trumps a general one); IOWA CODE § 4.7 (same); M INN. STAT . § 645.26 (same); NORTH DAKOTA CENTURY CODE

§1-02-07 (same); 1 PA. CON. STAT . §1933 (same); OHIO REV. CODE § 1.51 (same); TEXAS GOV’T CODE 311.026 (same); N.Y.MCKINNEY'S STAT . §238 (“Whenever there is a general and a particular provision in the same statute, the general doesnot overrule the particular but applies only where the particular enactment is inapplicable.”)

14

than specific ones -- his approach would maximize the satisfaction of political preferences.25 Whether hispremise is true is another matter, but that is more an empirical argument about what those preferences arethan a normative dispute about whether to maximize their satisfaction.

In fact, the empirical accuracy of Dworkin’s premise seems dubious. Politicians probably have farstronger convictions about concrete results than about abstract principles, and where the two conflict wouldprefer modification of the latter. Consider the case where a politician believes in the general principle ofpreserving species even at great national sacrifice, but an interpreter relies instead on the legislator’s morespecific conviction in completing a dam project even though it will endanger a species.26 Dworkin calls thisinterpretation a “crude . . . mistake, which no one would be tempted to make,” because this legislator hasmade a “mistake of fact” about whether the dam will endanger a species, and thus the interpreter shouldapply the general principle with the corrected facts.27 But it is more plausible that the legislator made a“mistake of principle” or that the interpreter made a mistake in what principle to attribute to the legislator.The legislator may have accepted an abstract principle without realizing all its consequences, but now thatshe realizes it would entail a concrete consequence she knows she finds unacceptable, she would modifythe general principle. Likewise, where a legislator believes that “reasonable” efforts should be made topreserve a species, is it a “mistake” for an interpreter to rely on the legislator’s specific belief that stoppingthis dam would be reasonable rather than on the interpreter’s own assessment that in fact stopping the damwould be unreasonable?28 To Dworkin it would be because this legislator has made a mistake “in fact”about the reasonableness of the dam,29 but one might well instead say that what must be mistaken (andmodified) are any general abstract criteria of reasonableness that failed to fit the legislator’s specificconcrete conclusion about this dam’s reasonableness.

To test this proposition, I searched the interpretive codes provided by legislatures within the UnitedStates to find out what they specified should govern in any conflict between general and specific legislativeconclusions. I discovered that every legislature to speak on the issue indicated that they preferred to havecourts to follow their more specific intent and provisions rather than their general intent and principles.30

I was unable to find any legislative statute directing courts to instead rely on evidence of the legislature’sgeneral intent. This seems the best available evidence on which default rule legislatures prefer. This defaultrule is also consistent with U.S. Supreme Court precedent, which adopts the anti-Dworkinian presumption

31 Morton v. Mancari, 417 U.S. 535, 550-51 (1974) (ruling that general statute against employment discriminationshould yield to earlier specific statute allowing Indian preferences) (collecting cases).

32 Elhauge, Interest Group Theory, supra note , at ___ (summarizing literature).33 Id. at __ (summarizing literature).34 Both conservative and liberal judges knowledgeable about the legislative process have accepted this point.

See Heath v. Varity Corp., 71 F.3d 256, 258 (7th Cir. 1995) (Easterbrook, J.); Boise Cascade Corp. v. FTC, 837 F.2d 1127,1158-59 (D.C. Cir. 1988) (Mikva, J., dissenting).

35 See Sunstein, Interpreting Statutes, supra note , at 426.36 See generally Elhauge, Preference-Eliciting, supra note , at X (discussing supplemental default rule that

operate where legislative preferences do not lend themselves to any reliable estimate).

15

that a conflict between specific and general legislation should be resolved in favor of the specific.31

Dworkin’s preference for abstract general principles over concrete specific legislative conclusionsthus, as he himself applies it, seems to both deviate from actual legislative preferences and allow judges tointerject their own value-laden judgments. But the problem is not limited to his particular methodology fordetermining statutory coherence. Legislation is often produced by political compromise or interest grouppressures that do not reflect a common policy.32 Even without interest group influence, Arrow’s Theoremshows that collective choices can make decisions that do not make a rational pattern.33 And even withouteither problem, separate enactments can simply reflect different concrete political reactions to distinctevents at varying points in time that do not fit into one overarching policy. All this means a series of statuteswill often reflect legislative preferences that fail to manifest a single coherent policy.34 More important,efforts to impose such a coherent policy on a set of enactments will inevitably require judicial valuejudgments about which reading makes the statutes work “best” together.35

None of this means it is irrelevant whether various pieces of legislation form a coherent policy. Allother things being equal, presumably legislators normally prefer a coherent policy to one that is not.36 Butthe issue of statutory coherence must be subordinated to the general issue of what maximizes politicalsatisfaction. A judicial or scholarly preference for coherence should not be permitted to trump evidencethat legislators have a different view about the desirability of coherence, or are more willing to trade it offagainst other goals. Nor, if legislators do prefer coherence, should their standards for what constitutes acoherent policy be replaced by different judicial or scholarly standards on policy coherence. Moregenerally, coherence arguments do not justify resolving statutory ambiguities in a way that conflicts withconclusions about what, given all the evidence (including coherence with other statutes), is most likely tomaximize the satisfaction of enactable preferences. In short, while statutory coherence is sometimesrelevant to determining what maximizes political satisfaction, the latter remains the ultimate standard andshould determine both what counts as coherence and how much coherence counts.

One can generalize this response to all sorts of related claims that might be made, like that statutoryinterpretation should advance the goal of stability or certainty or reliance. There are two possibilitiesregarding such claims. One is that they reflect norms that would maximize political satisfaction. If so, theseclaims offer no grounds for deviating from the approach offered here. The other possibility is that theydeviate from the norms that maximize political satisfaction. If so, my argument remains that judges haveno warrant (absent a constitutional limitation) for imposing substantive norms that increase political

37 See Elhauge, Preference-Eliciting, supra note , at XI.B (rejecting claims that reliance should trump politicalpreference satisfaction, but explaining the extent to which political satisfaction would be advanced by interpretationsthat are sufficiently stable to induce some behavioral reliance).

38 On the civic republican theory favoring deliberation, see generally Symposium: The Republican CivicTradition, 97 YALE L.J. 1493-1723 (1988); Frank Michelman, The Supreme Court, 1985 Term — Foreword: Traces of Self-Government, 100 HARV. L. REV. 4 (1986).

39 See Cass R. Sunstein, Deliberative Trouble? Why Groups Go to Extremes, 110 YALE L.J. 71 (2000); Schkade,Sunstein, Kahneman, Deliberating about Dollars: The Severity Shift, 100 COLUM. L. REV. 1139 (2000).

40 See Waldron, Legislators' Intentions and Unintentional Legislation, in LAW AND INTERPRETATION 329, 344-46(Andrei Marmor ed., 1995); Vermeule, Interpretive Choice, supra note , at 117.

41 Those favoring deliberation should favor preference-eliciting default rules even more strongly since theyencourage explicit deliberation. See generally Elhauge, Preference-Eliciting, supra note .

16

dissatisfaction. Norms like stability or certainty or reliance should not be permitted to trump the goal ofpolitical satisfaction, but should be considered to the extent relevant to determining how best to further thatgoal.37

C. Preference Satisfaction v. Fostering Deliberation. Finally, one might object that what judges should do is foster legislative deliberation, not preference

satisfaction.38 This could form the basis for its own alternative statutory default rule, but in practice isnormally invoked at the margins to justify certain exercises of judicial judgment over others.

A primary focus on political deliberation in my view overestimates both its capacity to moldpreferences and the likelihood that any molding will improve rather than worsen preferences. Whileindividual preferences are no doubt molded by the surrounding society, governmental efforts to moldpreferences are less successful, and the molding of preferences through political discussion is rarer still.

Even when it occurs, molding preferences through deliberation can often not be salutary. Politicaldiscussion frequently divides rather than joins us in seeing a common view of the public interest. Considercommon laments about the mean-spirited nature of political discourse -- or, if you are an academic,consider certain faculty meetings. Even when deliberation unites, that may be harmful because socialdynamics often lead people to become more polarized. Indeed, one of the initial proponents ofdeliberation-based theory, Cass Sunstein, has acknowledged that empirical tests show deliberation makesgroups more erratic and extreme in their judgments.39 Aggregating independent votes not only producesresults that are more consistent and moderate, but (assuming each voter is more likely than not to be right)their collective judgment is more accurate the greater the number of independent votes aggregated.40

In any event, a focus on deliberation is not inconsistent with the approach taken here, for myanalysis does not depend on any assumption that the preferences being satisfied are exogenous to theprocess of political deliberation. If deliberation helps reveal truer or deeper preferences, it is likely toproduce more ultimate preference satisfaction. They should also favor reliance on deliberated politicalpreferences as indicia of what the enacting or current legislature would have wanted.41 A deliberativemodel is thus broadly consistent with the approach taken here, although the terminology differs as wouldapplication.

D. At Least An Important Goal.

42ARISTOTLE, THE NICHOMACHEAN ETHICS, bk. 5, ch. 10 (W.D. Ross, trans., revised by J.O.Urmson, rev. Oxforded. 1984).

43 Burnet v. Guggenheim, 288 U.S. 280, 285 (1933).44 United States v. Klinger, 199 F.2d 645, 648 (2d Cir. 1952), aff'd per curiam, 345 U.S. 979 (1953).45 RICHARD POSNER, PROBLEMS OF JURISPRUDENCE 270 (1990). Posner has applied this view in his judicial

opinions. See Friedrich v. City of Chicago, 888 F.2d 511, 514 (7th Cir. 1989).

17

Suppose you are unconvinced by this Section, and persist in thinking that judicial judgment,statutory coherence, or fostering deliberation should play an important role in resolving statutoryambiguities. Even then, what follows still matters. After all, it is rare to see the interpretive theory thatdenies governmental preferences should play some role in resolving statutory ambiguities. Those who admitany weight to governmental preferences should thus be interested in seeing what it would look like todevelop statutory default rules that maximize those preferences. Those default rules would not, to suchpersons, be dispositive, but they would merit whatever weight such persons regard as appropriate to givegovernmental preferences within their own interpretive framework. To those who are convinced by thisSection, the default rules that follow will be even more important – resolving all the indeterminancies wherestatutory meaning and constitutional constraints do not bind.

III. ENACTING PREFERENCE-ESTIMATING STATUTORY DEFAULT RULES

Where statutory meaning is unclear, and provoking explicit legislative action is unrealistic, then thedefault rule that maximizes enactable political preferences will track the interpreter’s best estimate of whatthe legislative government would have chosen given its political preferences had they thought about theissue. This core concept has a long and distinguished pedigree. Aristotle himself urged that when faced witha statutory gap, the interpreter’s role is “to correct the omission -- to say what the legislator would havesaid had be been present, and would have put into law if he had known.”42 Justice Cardozo stated that toresolve a statutory ambiguity, “we have therefore to put to ourselves the question, Which choice is it themore likely that Congress would have made?”43 Judge Learned Hand likewise reasoned “what we do, andmust do, is to project ourselves as best we can, into the position of those who uttered the words, and toimpute to them how they would have dealt with the concrete occasion."44 More recently, Judge Posnerconcluded, “judges should ask themselves, when the message imparted by a statute is unclear, what thelegislature would have wanted them to do in such a case of failed communication.”45

But because these illustrious authors and judges all focused on what they presumed was a statute-specific inquiry, none focused on what general default rules to adopt about: (1) what one can assume anylegislature would want, (2) how to determine what specific legislatures would want, and (3) whichlegislature (the enacting or current legislature) to ask the question about. Their approaches also seemproblematic in the many cases where there is no one interpretation one can say is more likely than not whatthe legislature would have wanted. Framing the issue as a default rule helps deal with this problem, and

46 In some cases, there are reasons to think distributional effects or externalities are not priced into the contractor securities, generally because of informational problems or because third parties are affected. In those cases, there isgood reason not to adopt the default rule that maximizes the economic pie of the contracting parties. But absent somespecial reason like this, the default rule on default rules is to adopt the one that maximizes the total economic pie for thosewho form contracts and corporate charters.

47 There may be some question about whether the contracting parties’ preferences have changed over time.This likely explains the different contractual rules for long-term contracts and contracts about life-changing events, likesurrogate motherhood. But most contract cases are not of this sort.

18

with the more general problem that the legislature often has no common intent or purpose or desire, andestimates of what it would have enacted are often highly uncertain. Our task will be to choose those defaultrules that minimize the dissatisfaction of enactable preferences, not to pretend that there is one clearlegislative intent or purpose or desire when there is none or, rather, there are multiple ones.

The contracts and corporate literature has done far more to develop a series of specific defaultrules. But we immediately come to a big difference between statutes and business contracts or corporatecharters -- one that indicates that, while the general default rule framework of contract and corporate lawis useful, particular conclusions under it often are not. For business contracts and corporations, the normalassumption is that the participants would prefer the default rule that maximizes the economic pie on theassumption that any distributional effects would be reflected in (and thus offset by) the price of the contractor corporate securities.46 This permits the general assumption that all parties share a preference for themost efficient default rules, though to be sure which default rules are efficient may turn on personalcharacteristics of the parties such as their aversion to risk. With statutes, we have no warrant for assumingthat the legislative participants share the same set of preferences, and certainly no general grounds forassuming they prefer the most efficient default rules. One point of the political process is precisely to decidehow much to pursue efficiency versus other social goals. We thus need to inquire into the set of politicalpreferences possessed by a particular set of legislative participants to devise the appropriate default rules.In short, compared to corporate and contract law default rules, statutory default rules are more likely tobe tailored to the preferences of the particular set of legislative participants in question. The key questionwill thus be what default rules to adopt about how to do such tailoring, and how to deal with the inevitableuncertainty about what preferences are enactable.

For non-business contracts, people may care about things other than their economic gain, sochoosing efficient default rules is more questionable, and the contracting party preferences about defaultrules may show a variability similar to legislative polities. But now we come to a second big differencebetween contracts and statutes. Persons normally are not bound by old contracts they did not enter, soin contract law there is no question that the contracting parties would want the default rule that tracks theirown preferences.47 In contrast, governments represent constituencies that are governed not only by thestatutes they enact but also (indeed mainly) by old statutes enacted by prior governments. Statutoryanalysis (unlike contracts) thus raises the question: should courts track the preferences of the enacting orcurrent government?

For purposes of this Section, however, I will assume that the relevant governmental preferencesare static, or at least that the court has no reliable indication that they have changed from the time of

48 See Brogan v. United States, 522 U.S. 398, 406 (1998) (Scalia opinion for the Court); Frank H. Easterbrook, TheCase of the Speluncean Explorers: Revisited, 112 Harv. L. Rev. 1913, 1913-14 (1999) [hereinafter Easterbrook, SpelunceanExplorers]; Manning, Textualism and Equity, supra note , at 113-14. See also United States v. Kirby, 74 U.S. (7 Wall.)482, 485-87 (1868) (crime of willfully interfering with the mail does not apply to a police officer who is arresting a mailcarrier). But see Alex Kozinski, The Case of the Speluncean Explorers: Revisited, 112 Harv. L. Rev. 1876 (1999) (decliningto recognize any gap justifying an exception).

49 See Rotella v. Wood, 120 S. Ct. 1075, 1084 (2000) (opinion joined by Scalia & Thomas, JJ.); Manning,Textualism and Equity, supra note , at 114 & n.449 (collecting cases).

50 Easterbrook, Speluncean Explorers, supra note , at 1913-14; Manning, Textualism and Equity, supra note, at 113-15.

19

enactment. As we will see, a default rule based on static governmental preferences is not necessarily astatic default rule. For even static preferences may require updating a default rule to take into accountchanged circumstances. But putting aside the possibility that the governmental preferences themselves havechanged allows us to isolate the issues raised in ascertaining any government’s legislative preferences.

A. Universal Default Rules Fitting Any Governmental PreferenceStatutory default rules generally need to be tailored for specific legislative polities. But many canons

of statutory construction are general default rules that reflect what any legislative polity would want, giventhe plausible range of political preferences.

Many universal default rules cover particular recurring situations. For example, criminal statutesgenerally contain unqualified prohibitions on misconduct like taking another’s life, and do not specify whatto do when the person taking another’s life is a policeman stopping a terrorist or assassin from killing others.Even strong textualists do not conclude that the statute contains no exception and then put the policemanon trial for murder. Instead, they and courts generally treat all such criminal statutes as having a gap thatis filled with the doctrine that all criminal statutes include the defense of justification or necessity or self-defense.48 Likewise, statutes of limitations typically fail to specify what to do when the defendant hasfraudulently concealed her wrongdoing in a way that disabled the plaintiff from suing within the limitationsperiod. Again, even strong textualists treat the statute as having a gap, which they and the courts fill withthe doctrine that any statute of limitations is tolled during a period of fraudulent concealment.49

The textualist defense of such cases is generally that the legislature acts with the understanding thatthere exist background rules of interpretation like these exceptions.50 But the truth is that these rulings donot rest on any evidence that the enacting legislature actually knew about these background interpretiverules, and in any event this justification begs the question of which background interpretive rules the courtsshould choose. Indeed, this doctrinal defense would suggest that courts are free to adopt any interpretiverule they want as long as they publicize it sufficiently and make it generally applicable to all statutes.Understanding these doctrines as preference-estimating default rules provides a more persuasivejustification and limitation on the creation of such doctrines. These doctrines are appropriate default rulesonly because courts can be confident that, if any legislature did think about the issue, it would prefer notto criminally punish those trying to stop others from committing crimes or let off defendants because theyfraudulently concealed their wrongdoing. Doctrines that do not conform to legislative preferences could

51 They should accordingly be adopted only in the limited conditions justifying a preference-eliciting defaultrule. See Elhauge, Preference-Eliciting, supra note .

52 See, e.g, United States v. X-Citement Video, Inc., 513 U.S. 64, 68-69 (1994); Public Citizen v. United States, 491U.S. 440, 454 (1989); Church of Holy Trinity v. United States, 143 U.S. 457, 459 (1892). These cases suggest that thisdoctrine also has a hermeneutic counterpart: that absurdity is a reason for finding ambiguous a meaning that wouldotherwise follow from a “literal” or “most natural grammatical reading” of a statute. I leave that hermeneutic questionto one side since my focus here is on the default rule that applies once ambiguity is admitted. See generally Manning,Textualism and Equity, supra note , at 55 n.227, 113-19 (collecting conflicting authority on whether plain meaninggoverns even when absurd).

53 United States v. Kirby, 74 U.S. 482, 487 (1868); 1 WILLIAM BLACKSTONE, COMMENTARIES *60.54 United States v. Ron Pair Enterprises, 489 U.S. 235, 242 (1989) (quoting Griffin v. Oceanic Contractors, Inc.,

458 U.S. 564, 571 (1982)); Mova Pharmaceutical Corp. v. Shalala, 140 F.3d 1060, 1068 (D.C. Cir.1998). Again, I am limitingmy analysis to cases where the statutory meaning itself is ambiguous. For an excellent summary of the hermeneuticdebate on whether a statutory purpose should trump an otherwise clear statutory text, see Manning, Textualism andEquity, supra note , at 3-7, 9-22.

55 Sturges v. Crowninshield, 17 U.S. (4 Wheat.) 122, 202-03 (1819) (Marshall, C.J.).

20

also form “background” interpretive rules, but cannot be justified as preference-estimating default rules.51

All the above examples could be understood as specific instances of the more general canon thatstatutes should be interpreted to avoid absurd results.52 To take a famous example, a statute in Bolognathat forbids residents to “draw blood in the streets” is read not to prohibit a surgeon from drawing bloodin a medical emergency but rather to prohibit swordplay.53 An interpretation is considered absurd not onlywhen it violates common sense, but also when it conflicts with the statute’s purpose.54 The latter presumesthat an ascertainable statutory purpose, which alas is not always the case.

The absurdity canon is sometimes criticized on the ground that it invites judges to substitute theirown preferences for statutory meaning. Using the canon to impose judicial preferences would certainly beimproper. But the canon is more commonly applied to cases like the Bologna statute, where commonsense indicates that any government would prefer to avoid a certain statutory result. In such a case, thecourt can feel confident that rejecting this interpretation furthers enactable preferences, even though thecourt has no concrete indication that the particular enactors ever thought about the matter or expressedpreferences on it. Explicitly supporting this notion was Justice Marshall’s test that the absurdity must be“so monstrous, that all mankind would, without hesitation, unite in rejecting the application.”55

In other cases, the interpretation rejected as “absurd” does not conflict with common sense, butrather with the preferences reflected in the general statutory purpose. But what are judicial findings of“purpose’? Although courts use the term in varying ways, the most common usage is restricted to apurpose inferred from the general statute that contains the provision being interpreted. Under myterminology, this can be seen as a special case of a preference-estimating default rule: the case where theenactable preferences can be estimated from the best possible source of information, the actual enactmentof the statute.

In either case, the justification for rejecting the absurd interpretation is not a conflict with judicialpreferences. An “absurd” interpretation should thus not be understood to mean an interpretation that thejudge believes reflects a highly unwise policy. It is rather a statutory result that the court is confident either

56 HAWAII REV. STAT . § 1-15 (“Where the words of a law are ambiguous: . . . Every construction which leads toan absurdity shall be rejected.”); LOUISIANA CIVIL CODE Art. 9 (“When a law is clear and unambiguous and its applicationdoes not lead to absurd consequences, the law shall be applied as written”); M INN. STAT . § 645.17 (in construingstatutes, courts can presume that the legislature “does not intend a result that is absurd”; 1 PA. CON. STAT . §1922 (same); NEW M EXICO STAT ANN, 1978, § 12-2A-18 (“A statute or rule is construed, if possible, to: . . . avoid an . . . absurd . . .result.”); N.Y. MCKINNEY'S STAT . §145 (“A construction which would make a statute absurd will be rejected.”).

57 See, e.g., Solid Waste Agency of N. Cook County v. U.S. Army Corps of Eng'rs, 121 S. Ct. 675, 683 (2001);Public Citizen v. DOJ, 491 U.S. 440 (1989); Edward J. DeBartolo Corp. v. Fla. Gulf Coast Bldg. & Constr. Trades Council,485 U.S. 568, 574-75 (1988); NLRB v. Catholic Bishop, 440 U.S. 490 (1979).

58 See Elhauge, Preference-Eliciting, supra note , at X, which also notes observes that some constitutional normsmight justifiably be deemed “under-enforced” and thus justify application of this canon on constitutional rather thaninterpretive grounds. One might regard this canon as a preference-eliciting default rule, but it would be a poor one forthere is no reason to think that all groups who secure the enactment of laws that raise constitutional difficulties can re-gain access to the legislative agenda to override adverse interpretations. See id.

21

that the actual enacting government would not want or that no government (within the range of plausiblepolitical preferences in that jurisdiction) would ever want.

One might object that, although in theory a canon against absurd interpretations would advancepolitical satisfaction, in practice judges would abuse the discretion conferred by that canon in ways thatthwart legislative preferences. If so, one might expect to see legislatures opting out of this default rule. Butin fact every legislature within the United States has enacted a code of construction that leaves the well-known canon against absurd interpretations undisturbed. Further, those legislatures that have consideredthe matter expressly have all adopted the canon against absurd interpretations.56 No legislature directscourts to follow plain meaning of statute even when it leads to absurd results. This evidence tends toconfirm the analysis above that this canon reflects a general default rule that maximizes political satisfaction.

Another example of a universal default rule is the canon against interpreting statutes to createconstitutional invalidity. Again, any government would want such a default because it would want topreserve as much of its statute as possible to maximize the satisfaction of the political preferences that ledto the enactment. Likewise, it would want a canon of severability to preserve as much of any statute thatwas partially invalidated as can be preserved without undermining the statutory scheme.

The canon interpreting statutes to avoid constitutional questions or difficulties,57 on the other hand,cannot be explained as a preference-estimating default rule unless it is confined to cases where the courthas no idea what legislative preferences are most likely for the case at hand.58 Otherwise, in the caseswhere it has any bite, this default rule would reject an interpretation that likely reflects governmentalpreferences even though that interpretation is not, in fact, constitutionally invalid. (It has no bite in casesof actual constitutional invalidity because the first default rule already governs those cases). This is not likelyto reflect the preferences of any government, which prefers to maximize the satisfaction of its politicalpreferences.

Again, the evidence from the actual legislative codes of statutory interpretation are consistent withthis analysis. Every legislature in the United States to enact a statute of interpretation on the subject has

59 COL. REV. STAT . §§ 2-4-201; IOWA CODE § 4.4; M INN. STAT . § 645.17; 1 PA. CON. STAT . §1922; NEW MEXICO

STAT ANN, 1978, § 12-2A-18; NORTH DAKOTA CENTURY CODE §1-02-38; OHIO REV. CODE § 1.47; TEXAS GOV’T CODE

311.021. Every state legislature appears to have a provision providing for severability.60 See FALLON, MELTZER & SHAPIRO, supra note , at 829, 839.61 Id. at 841-42. Other arguments take the position that the courts should not fill this gap in order to force the

legislature to decide the question. Id. at 842-43. But since the question whether to provide a private right of action isunlikely to involve one-sided political demand for legislative correction of mistaken judicial estimates, this is a poor topicfor what (if this argument were credited) amounts to a preference-eliciting default rule. See Elhauge, Preference-Eliciting,supra note , at Part II.

62 Id. at 840-41.63 Id. at 840 & n.4, 843-44 (noting that recent cases rely on the specific legislative intent indicated by the text

or legislative history rather than a general default rule one way or the other).64 Against the use of legislative history to determine statutory meaning, see, e.g., Conroy v. Aniskof, 113 S. Ct.

1562, 1567-72 (1993) (Scalia, J., concurring in the judgment); REED DICKERSON , THE INTERPRETATION AND APPLICATION

OF STATUTES 155-58, 174-75 (1975); Adrian Vermeule, Legislative History and the Limits of Judicial Competence, 50STAN. L. REV. 1833 (1998) [hereinafter Vermeule, Legislative History]; W. David Slawson, Legislative History and the

22

directed courts to construe statutes to avoid constitutional invalidity.59 I was unable to find any statute thatalso directed courts to avoid constitutional doubts that did not result in actual invalidity.

Given the variety of political preferences, the occasions for universal default rules are rather limited.Consider, for example, a statute that defines a violation without specifying whether or not it means toauthorize a private remedy for violations. At one point, under the slogan that every right must have aremedy, courts applied the general default rule of providing of a private remedy since that would furtherthe statutory purpose of deterring the violation.60 But this one-sided understanding of legislative preferencesproperly came under critique. Legislative preferences may reflect a compromise between competingpolitical interests, a tradeoff that balances a desire for enforcement against concerns about overdeterringdesirable conduct, or a belief that agencies can be trusted with enforcement discretion better than privateparties.61 This briefly led courts to the opposite default rule that the legislature does not mean to providefor a private right of action.62 But (if one accepts the premise that the statute is indeterminate) this too isincorrect. Rather the correct conclusion to draw is that enactable preferences on the topic are too variableto assume any sort of universal legislative preference exists. Instead, resolving the statutory indeterminancyrequires examination of the specific preferences of the legislature in question, which corresponds to theposition on which modern law has finally settled.63

The question whether to imply private rights of actions is just one illustration of the generalproposition that the default rules for resolving statutory indeterminacy normally must be tailored to take intoaccount the variety of governmental preferences. I turn next to the topic of how to conduct such a tailoredanalysis.

B. Tailored Default Rules Reflecting Legislative HistoryIn discussing how courts should tailor default rules to inquire into what the enacting legislature

would have wanted, we risk treading on hermeneutic theory, and in particular a voluminous debate aboutwhether legislative history is or is not an accurate reflection of statutory meaning.64 Let me avoid this thicket

Need to Bring Statutory Interpretation Under the Rule of Law, 4 4 STANFORD L. REV. 383 (1992); Jerry L. Mashaw,Textualism, Constitutionalism, and the Interpretation of Federal Statutes, 32 Wm. & Mary L. Rev. 827 (1991); FrankH. Easterbrook, Text, History, and Structure in Statutory Interpretation, 17 HARV. J.L. & PUB. POL'Y 61 (1994) [hereinafterEasterbrook, Text]; Frederick Schauer, Statutory Construction and the Coordinating Function of Plain Meaning, 1990S. CT. REV. 231; Frank Easterbrook, Role of Original Intent in Statutory Construction, 11 HARV. J.L. & PUB. POL. 59(1988); Kenneth W. Starr, Observations about the Use of Legislative History 1987 DUKE L.J. 371; Max Radin, StatutoryInterpretation, 43 HARV. L. REV. 863, 870-71 (1930).

In favor of using legislative history, see Stephen F. Ross & Daniel Tranen, The Modern Parol Evidence Ruleand Its Implications for New Textualist Statutory Interpretation, 87 Georgetown L. J. 195 (1998); Cass R. Sunstein,Justice Scalia’s Democratic Formalism, 107 YALE L.J. 529 (1997); Stephen Breyer, On the Uses of Legislative Historyin Interpreting Statutes, 65 S. CAL. L. REV. 845 (1990); George Costello, Average Voting Members and Other BenignFictions, 1990 DUKE L. J. 39; Daniel A. Farber & Philip Frickey, Legislative Intent and Public Choice, 74 VA. L. REV. 423(1988); Abner Mikva, A Reply to Judge Starr’s Observations, 1987 DUKE L. J. 380; Patricia M. Wald, Some Observationson the Use of Legislative History in the 1981 Supreme Court Term, 68 IOWA L. REV. 195 (1983); James M. Landis, A Noteon Statutory Interpretation, 43 HARV. L. REV. 886, 888-89 (1930).

65 See, e.g., Radin, supra note, at 871; sources cites supra note.66 See, e.g., N.Y. M CK INNEY’S STATUTES §§124-25 (“proper to consider the legislative history of the act, the

circumstances surrounding the statute's passage, and the history of the times. . . If the interpretation to be attached to

23

by simply saying this: whether or not one invites legislative history into the initial inquiry as to a statute’smeaning, a separate question is whether legislative history should be used to resolve indeterminancy whenthe interpreter is left genuinely uncertain as to the statute’s meaning after employing her favorite hermeneutictechniques. Used not to divine what the statute actually meant, nor what legislature subjectively intended,as to the specific matter in question. Used rather, in cases where the interpretive inquiry is inconclusive,to make frankly probabilistic estimates about which statutory interpretations are most likely to reflectenactable political preferences.

My answer is that it should. Legislative history may not alter statutory meaning under certainhermeneutic theories that stress that what the legislature enacts is the statutory text, not the legislativehistory.65 But legislative history certainly helps estimate the general political preferences that motivated theenacting government at that time. And those estimates of political preferences are all we have, wherestatutory meaning is unclear, to determine which interpretation is least likely to produce politicaldissatisfaction. The enacting government itself would want a default rule whereby courts refer to legislativehistory to resolve gaps and ambiguities in its meaning because that increases the extent to which its politicalpreferences are satisfied by statutory interpretations.

We might test this proposition by inquiring about what legislatures have actually provided in theircodes of construction. Since it has long been the practice of judges to inquire into legislative history, onewould think that if this practice increased political dissatisfaction, then legislatures would try to outlaw itwhen they enact codes of statutory construction. In fact, although each state legislature and Congress hasenacted such a construction code, none has outlawed reliance on legislative history. We should hesitatebefore relying too heavily on this fact standing alone, because the United States and many states haveenacted construction codes that are largely inconsequential, and thus may have never considered the issue.But there is more telling evidence. The ten state legislatures to enact statutes that did consider the issue allexpressly authorize courts to look at legislative history.66 Further, thirty other state legislatures have

a statute is doubtful, the courts may utilize legislative proceedings to ascertain the legislative intent.”); COL. REV. STAT .§§ 2-4-203 (“If a statute is ambiguous, the court . . . may consider . . . The circumstances under which the statute wasenacted” and “The legislative history”); IOWA CODE § 4.6 (same); NORTH DAKOTA CENTURY CODE §1-02-39 (same); OHIO

REV. CODE § 1.49 (same); M INN. STAT . § 645.16 (“When the words of a law are not explicit, the intention of the legislaturemay be ascertained by considering . . . The circumstances under which it was enacted” and “The contemporaneouslegislat ive history”); 1 PA. CON. STAT . §1921 (same); N.J. STAT . 2A:58C-1 (“The Legislature further finds that suchsponsors' or committee statements that may be adopted or included in the legislative history of this act shall beconsulted in the interpretation and construction of this act.”); NEW MEXICO STAT ANN, 1978, §12-2A-20 (“If . . . themeaning of the text or its application is uncertain, the following aids to construction may be considered in ascertainingthe meaning of the text: (1) the circumstances that prompted the enactment or adoption of the statute or rule; (2) thepurpose of a statute or rule as determined from the legislative or administrative history of the statute or rule; and (3) thehistory of other legislation on the same subject.”); TEXAS GOV’T CODE 311.023 (“In construing a statute, whether or notthe statute is considered ambiguous on its face, a court may consider . . .the . . . circumstances under which the statutewas enacted” and “legislative history”).

67 ARIZ .REV. STAT . § 1-211 (“objects”); ARKANSAS CODE § 1-2-202 (“true intent”); CAL. CODE CIV. PROC. §1859(“intention”); 1 DELAWARE CODE § 301 (“intent”); GEORGIA CODE §1-3-1 (“intention”); HAWAII REV. STAT . § 1-15 (“Thereason and spirit of the law, and the cause which induced the legislature to enact it”); IDAHO CODE §73-102 (“objects”);5 ILLINOIS COMPIL. STAT . 70/1.01 (“true intent”); INDIANA CODE §§1-1-4-1, 1-1-4-5 (“inten t” ) ; KANSAS STAT . § 77-201(“intent”); KENTUCKY REV. STAT . §446.080 (“objects” and “intent”); LOUISIANA CIVIL CODE Art. 10 (“purpose”); MASS.GEN. LAW. ch. 4 §6-7 (“intent” and “intention”); M ICHIGAN COMPLIED LAW §8.3 (“intent”); M ISS. CODE ANN. § 1-3-1(“object”); M ISSOURI STAT . §1.010 (“true intent”); MONTANA CODE §§ 1-2-102, 1-2-103 (“intention” and “objects”);NEBRASKA REV. STAT . §§ 49-802, 49-806 (“intent”); NEW HAMPSHIRE REV. STAT . § 21:1 (“intent”); N.J. STAT . 1:1-1(“intent”); NORTH CAROLINA GEN. STAT . § 12-3 (“intent”); 25 OK. ST. ANN. § 29 (“objects”); OREGON REV. STAT . § 174.020(“intention”); RHODE ISLAND STAT . § 43-3-2 (“intent”); SOUTH DAKOTA CODIFIED LAW §2-14-12 (“objects”); UTAH CODE

1953 § 68-3-2 (“objects”); 1 VERMONT STAT . § 101 (“intent”); VA. CODE ANN. §§ 1-13, 18.2-5, 19.2-5 (“intention” and“intent”); W.VA. STAT . § 2-2-10 (“intent”); WISCONSIN STAT . §§ 990.001, 990.01 (“intent”). In addition, the Washingtonlegislature enacted a statute directing that its statutes be “liberally construed,” WASHINGTON STAT §1.12.010, whichprobably was intended to authorize inquiries into legislative purposes and has been so read by the Washington courts.I include the New Jersey statute here because the statute cited in the preceding footnote concerned the legislativehistory of a particular act, whereas the statute in this footnote is the one that generally applies to all statutoryconstruction. Since I counted it before, I omit it here to avoid double-counting.

68 There has, however, been one statute-specific opt-out discussed infra at __.

24

directed judges to consider legislative intent or purposes.67 Given that the prevailing methods forascertaining such intent and purpose included judicial inquiry into legislative history, these statutes wouldseem to implicitly authorize such inquiry. All told then, 40 of the 50 state legislatures have enacted statutesthat explicitly or implicitly authorize judicial inquiry into legislative history and intent. Coupled with the factthat no legislature has enacted a statute directing courts to generally ignore legislative history,68 this helpsconfirm the conclusion that it maximizes political satisfaction to adopt the general default rule that, wherethe statute is ambiguous, courts should examine legislative history in order to estimate legislativepreferences.

It might justifiably be doubted that any theory could explain the pattern of uses of legislative historyby the courts, which often seems highly inconsistent from case to case. But the distinction betweenhermeneutic and default rule theory may explain at least some of the seeming inconsistency. A judge mightignore legislative history in one case because she considers it irrelevant to understanding a statutory meaning

69 See McNollgast, Positive Canons, supra note , at 719-20, 724.70 See Manning, Textualism as Non-Delegation, supra note , at 733-34 & n.243 (collecting cases that rely on

such evidence when interpreting statutes); see also supra note __ (noting that state statutes that address the issuedirect courts to consider the circumstances that provoked the statute).

71 Keifer & Keifer v. Reconstruction Fin. Corp., 306 U.S. 381, 389 (1939) (statutory interpretation draws“significance from dominant contemporaneous opinion."); Archibald Cox, Judge Learned Hand and the Interpretationof Statutes, 60 Harv. L. Rev. 370, 370 (1947) ("[T]he courts have never had serious difficulty in using the history of thetimes as well as the committee reports and debates in Congress to ascertain the evil which the legislation was designedto remedy."); Felix Frankfurter, Some Reflections on the Reading of Statutes, 47 Colum. L. Rev. 527, 539 (1947) ("Statutescannot be read intelligently if the eye is closed to . . . the known temper of legislative opinion."); infra note (collectingstate statutes that direct courts to examine the “circumstances surrounding the statute's passage” and the “history ofthe times”).

72 See, e.g., Keifer, 306 U.S. at 389 ("The Congressional will must be divined, and by a process of interpretationwhich, in effect, is the ascertainment of policy immanent not merely in the single statute... but in a series of statutes");Frankfurter, supra note , at 539 (intelligent statutory interpretation requires examining “considerations evidenced in

25

that can independently be established. The same judge might then use legislative history in the next casebecause meaning is ambiguous and the history helps in estimating governmental preferences. It thus is notnecessarily inconsistent for a judge to reject legislative history in some cases but not others. Other judgesmay have no general aversion to using legislative history to figure out statutory meaning, but in one casemight decline to look at post-enactment legislative history because they consider that irrelevant tostatutory meaning. The same judges might then consistently consider such post-enactment legislativehistory in the next case because statutory meaning is unclear and the court finds that post-enactment historysufficiently close in time to help estimate the enacting governmental preferences and thus resolve theambiguous statutory meaning. Again, the mere fact that post-enactment legislative history is sometimesused and sometimes rejected does not prove inconsistency.

The phrase “legislative history” is sometimes read too literally to include only statements or actionsby legislators. But in systems like the U.S. that give executives the power to veto legislation, one mustconsider executive views since they strongly influence what preferences are actually enactable.69 Defaultrule analysis also suggests that, where justified by ambiguous statutory meaning, inquiry into likelygovernmental preferences should be broader than legislative or executive deliberations about the particularambiguity at issue. The inquiry should include any legislative history that helps estimate the generalenactable preferences of the government, whether or not that legislative history involves specific answersto the specific question of meaning. Indeed, the inquiry should – and in actual legal practice does – eveninclude general information about the events that provoked the enactment70 or about which political forcesare most influential during the particular historical period.71

It should also include legislative history regarding entirely different statutes. Such debates canprovide information bearing on those political preferences that are relevant to determining how the enactinggovernment would have wanted to resolve the statutory indeterminancy in question. Even more revealingare the pattern of statutes actually enacted before or after the statute in question, which indicates thepolitical preferences that can actually command legislative action. Again, this is often used in actualstatutory interpretation.72 Indeed, it provides an alternative basis for the statutory canon that, where

affiliated statutes”).73 See, e.g., FDA v. Brown & Williamson, 120 S. Ct. 1291, 1301 (2000); United States v. Monsanto, 491 U.S. 600

(1989); Public Employees Retirement System v. Betts, 492 U.S. 158 (1989); United States v. Fausto, 484 U.S. 439, 453(1988); see also DWORKIN, supra note , at 176-84, 190-92, 217, 225. Unlike Dworkin, however, the theory here (anddoctrine just cited) is supplementary: it would not justify interpreting a statute contrary to legislative preferences in orderto avoid an inconsistency that apparently did not bother the legislature. See also supra Section II.

74 NEW MEXICO ST A T ANN, 1978, §12-2A-20 (in construing a statute a court may consider “a statute or rule onthe same or a related subject, even if it was enacted or adopted at a different time” and where ambiguous “the historyof other legislation on the same subject”); HAWAII REV. STAT . § 1-16 (“Laws in pari materia, or upon the same subjectmatter, shall be construed with reference to each other.”); LOUISIANA CIVIL CODE Art. 13 (“Laws on the same subjectmatter must be interpreted in reference to each other.”); MINN. STAT . § 645.16 (in construing statutes, courts shouldconsider “other laws upon the same or similar subjects”); 1 PA. CON. STAT . §1921 (same); COL. REV. STAT . §§ 2-4-203 (same); IOWA CODE § 4.6 (same); NORTH DAKOTA CENTURY CODE §1-02-39 (same); OHIO REV. CODE § 1.49 (same); TEXAS

GOV’T CODE 311.023 (same); N.Y. MCKINNEY'S STAT . §221-23 (“The phrase ‘in pari materia’ is applied to statutes orgeneral laws, usually enacted at different times but with reference to the same subject matter. . . .A statute is to beconstrued with reference to earlier statutes in pari materia. . . A subsequent act in pari materia may be considered as anaid in the construction of an earlier statute or section.”).

75 Radin, supra note, at 870-71.76 See Easterbrook, Statutes’ Domains, 50 U. CHIC. L. REV. 533, 547-48 (1983); Kenneth A. Shepsle, Congress

is a 'They', Not an 'It': Legislative Intent as Oxymoron, 12 INT 'L REV. L. & ECON. 239 (1992); Manning, Textualism andEquity, supra note , at 19-20 (collecting sources). See generally Elhauge, Interest Group Theory, supra note , at 101(collecting and explaining literature on the legislative cycling problem and Arrow’s Theorem).

26

ambiguous, statutes should be interpreted to be consistent with other enactments on the same subject.73

Consistent with this analysis, every legislature to advert to the question has also agreed that its statutesshould be interpreted to be consistent with other statutes on the same subject whether enacted before orafter the statute in question.74 Legislative history of this form, that consists of actual legislative action, orwidespread political forces, is in fact likely to be a more accurate indicator of enactable preferences thanthose forms of legislative history that consist of statements by individual legislators or committees.

Framing the issue as a question of default rule, rather than of hermeneutics or legislative intent, helpsrespond to the three objections commonly made to the use of legislative history.

Objection 1. There Is No Legislative Intent. Max Radin long ago argued that it made no senseto speak of a common legislative intent or purpose when legislative action was the product of combinedaction by multiple actors, each of whom had their own motives.75 More recently, scholars have relied onrational choice theory to argue that legislative preferences often cycle, so that one cannot speak of whatthe legislature intended or would have wanted because its desires are often intransitive, so that what itchooses may depend solely on how its agenda is structured.76

These may well be valid objections to a hermeneutic theory that depends on the proposition thatinterpreters can determine with certainty one common legislative intent or purpose that would establish afixed statutory meaning. Radin himself assumed the need for great certainty, demanding evidence “severalhundred [legislators] each will have exactly the same determinate situations in mind” and that “the litigatedissue, will not only be within the minds of all these [legislators], but will be certain to be selected by all of

77 Radin, supra note, at 870.78 There might also be positive odds that legislative deadlock or preferences would have prevented either

interpretation from being enacted. In that event, though, minimizing the dissatisfaction of enactable preferences wouldinvolve figuring out the relative likelihood of each interpretive option being enacted if the government had made achoice. For example, if the first option is 30% likely to be what the legislature would have enacted, and the second optionis 20% likely, the first is more likely to minimize political dissatisfaction, and the relative probabilities (as I am using theterm) of enactability among the two options would b 60% and 40%. I defer until Part IV below the interesting case where,given three or more interpretive options, no single option has a relative probability of being accurate that is greater than50%.

Sometimes, the odds of deadlock might be so high that there is no relevant enactable preference for a particulargovernment. In that case, a court might shift its focus from enacting governmental preferences to current governmentalpreferences, or vice versa, or look at the most recent indication of enactable preferences. See infra at V-VII.. An inabilityto meaningfully estimate enactable preferences might also justify a supplemental or preference-eliciting default rule. SeeElhauge, Preference-Eliciting, supra note .

79 See Elhauge, Interest Group Theory, supra note , at 102 (collecting literature).80 See Schwartz, Statutory Interpretation, Capture, and Tort Law, 2 AMER. ECON. REV. 1, 13-15 (2000)

(summarizing literature).

27

them.”77 But statutory default rules need not depend on any such proposition. Instead, the only propositionI will be relying on is that interpreters can make some, perhaps rough, probabilistic assessment of therelative enactability of the various interpretive options that seem plausible after hermeneutic analysis.

Suppose, for example, a court concludes that there are two plausible interpretations and that therelative probabilities that the legislature would have enacted the interpretations given a choice between thetwo are 60% for first, and 40% for the second.78 The court need not be under any illusion that the firstinterpretation was in the minds of all the enacting legislators’, was dictated by some general purpose thatwas in their minds, or in any other sense reflected a common legislative intent or purpose. The court needmerely have estimated that, if the legislature had addressed the issue, the first interpretation is more likelyto have been what it would have enacted than the second. Assuming these probabilistic estimates are moreaccurate than random guesses, choosing the first interpretation would still minimize the dissatisfaction ofenactable preferences. Thus the enacting legislature would want courts to choose a default rule that trackedtheir best estimate of enactable legislative preferences even if it does not reflect their common legislative“intent” in any meaningful way.

But, one might object, the cycling problem means that one may not be able to meaningfully assess(or even order) the odds of various interpretations being enacted because the legislature would cycle amongthem. This objection overstates the problem. While Kenneth Arrow proved that it was impossible todefine a method of aggregating preferences that satisfied certain rationality principles and never cycled, thisdoes not mean that cycling always occurs. In fact, empirical literature has proven that (as commonobservation indicates) endless legislative cycling is not the norm.79 Indeed, on plausible assumptions, theodds that majority preferences will not cycle on any given policy choice are 94-99%.80 Further, if votingbehavior is probabilistic rather than known with certainty, then public choice theory itself shows that

81See Elhauge, Interest Group Theory, supra note , at 102 n.270; DENNIS MUELLER, PUBLIC CHOICE II 198-203(1989).

82 Elhauge, Interest Group Theory, supra note , at 102 (collecting lite ra ture) ; MUELLER, supra note , at __;McNollgast, Positive Canons, supra note , at .

83 See Elhauge, Interest Group Theory, supra note , at 104-05 (collecting sources).84 See Manning, Textualism as Non-Delegation, supra note , at 679-80 & nn. 22, 27.85 McNollgast, Positive Canons, supra note , at 725-26.86 Id. at 720-27. If the actors’ political preferences lay on both sides of the status quo, then no legislation could

be enacted because some actor would veto any shift from the status quo to a worse direction.87 The veto point for each actor is not its ideal point but the point at which the proposal has gotten so far away

from its ideal point that it would prefer the status quo and will thus veto the proposal. Id. at 722-24.88 When one actor requires the concurrence of a second actor to make its veto stick, like a President who needs

the political support of one-third the legislators to sustain his veto, their collective veto point will be the furthest of their

28

majority voting can avoid cycling problems.81 And a great body of theoretical work has shown how, evenwhen majority preferences are cyclical, equilibrium is commonly produced by legislative structures, like theneed to get committee approval and consent from multiple legislative houses and either the executive or alegislative super-majority.82

Even where cycling problems mean that legislative results depend on how the agenda is structured,the fact remains that legislative agendas are structured in ways that courts can take into account. Inparticular, the cycling literature indicates that what the legislature enacts can turn heavily on the politicalpreferences of whichever elected officials have agenda-setting power, like committee chairs, legislativesponsors, and legislative leaders.83 The political preferences of these agenda-setters are thus very likelyto influence where enactable preferences lie. Consistent with this, interpreting courts treats committeereports and sponsor statements as especially persuasive legislative history.84 Even without any cyclingissues, committee reports offer strong evidence about enactable preferences in any legislative system thateffectively requires committee consent to enact legislation within that committee’s jurisdiction. In particular,the fact that a committee report rejects a particular interpretation accurately indicates that such aninterpretation does not match enactable legislative preferences, either because the committee does notprefer that interpretation or because the committee has determined that it must signal opposition to thatinterpretation to procure enactment. As between two interpretive options, the one rejected by a committeewith veto power can thus safely be rejected as having no chance of reflecting enactable preferences.85

More generally, influential rational choice modelers like McNollgast have thus concluded that, whenone takes into account the actual existence of agenda and veto powers, one can arrive at determinantconclusions about whose political preferences determined what was enactable. In their model, they assumelegislation requires the consent of multiple actors (committees, house majorities, the President) who havepolitical preferences that lie in a policy spectrum on one side of the status quo.86 For any legislation thatoccurs, the preferences of the actor who can last propose a law that the others can only accept or vetowill govern if those preferences are closer to the status quo than the veto-point of any other actor.87

Otherwise, the political preferences that govern will be those of the actor with the veto point closest to thestatus quo.88 The legislative result thus depends only on where the political preferences lie in relation to the

two veto points from the status quo. Id.89 For example, if the legislative procedure does not allow amendments, then the committee is the last offeror

and has more influence. If the legislative procedure does allow amendments, then the house majority (perhapsrepresented by its majority leader or caucus) is the last offeror. Id. at 722-25.

90 See supra at II.A.91 See Manning, Textualism and Equity, supra note , at 18, 20-22 (collecting sources).92 See McNollgast, Positive Canons, supra note , at 705-07.

29

status quo and whom the legislative rules allow to make the last offer.89

There are problems with this model. In particular, it effectively assumes that the last offeror cangrab all the joint gains from agreement for itself. Thus, in the model Congress can offer the President astatute only marginally better than one he would regard as worse than the status quo. But in fact, aPresident is likely to veto such a statute not only in order to get a better deal on that statute but also toprevent Congress from persisting in such one-sided deals in the future. More generally, given that thelegislative participants are repeat players, they are instead likely to veto such one-sided deals in order toinduce offers that give them a larger share of the joint gains in the future. Thus, the McNollgast modelseems to overstate the certainty with which courts could determine whose political preferences wereenactable. But, coupled with some common sense about bargaining dynamics, it does provide helpfulguidance for at least making estimates about enactable preferences.

One might still object that this approach allows the result to be distorted by the veto or agenda-setting power of certain legislators who are not representative of the entire legislature. But under apreference-estimating default rule, the court’s role is not to determine what the legislature would haveenacted if it had a different legislative process than it actually has.90 That legislative process, whatever itsflaws, is apparently legitimate enough to justify compelled obedience to enactments whose meaning can bedivined. The default rule just aims to resolve the ambiguities that legislature left by minimizing thedissatisfaction of those preferences that are enactable under the regime that actually exists. Further, sincethe very literature that suggests great agenda-setting power also indicates some agenda-setting is necessaryto avoid cycling, it hardly makes sense to ask what the legislature would have done without anyone settingthe agenda. That question has no answer. Nor can one realistically speak of the legislative process being“distorted” by agenda setting power when it is inevitable that any process would necessarily have to havesomeone set the agenda.

Other textualist objections to inquiries into legislative purpose are that statutes may reflect a bargainor tradeoff, or may choose a rule over a standard, either of which could be upset by using generalizationsabout statutory purposes to extend the statute or to convert a statutory rule into a judicial standard.91 Theseare valid objections to loose-minded purposivism, but not to looking to legislative history or estimatinglegislative preferences. To the contrary, examining legislative history (capaciously defined) will often revealthe limits of the bargain92 or a preference for a rule, both of which are proper inquiries for estimatinglegislative preferences. But there is no reason to assume that in all cases legislatures meant to adopt themost narrow bargain or to choose bright-line rules over flexible standards.

In short, a preference-estimating default rule calls for estimating the probabilities of enactable

93 See Elhauge, Preference-Eliciting, supra note , at II (specifying conditions under which this will be true). 94 Id. at X.95 Easterbrook, Statutes’ Domains, supra note , at 548.96 See, e.g., Conroy v. Aniskof, 113 S. Ct. 1562, 1567-72 (1993) (Scalia, J., concurring in the judgment); Radin,

supra note, at 870-71; DICKERSON, supra note , at 155-58, 174-75; Vermeule, Legislative History, supra note , at 1896.97 See, e.g., Landis, supra note , at 888-89; sources cites supra note . In one sense, legislative history is likely

to be far more accurate than the extrinsic evidence commonly used to interpret ambiguous text or fill defaults in contractcases. Pre-contract statements are often oral or documented in only a fragmentary fashion, raising serious questionsabout the accuracy of any finding about what statements were made. For legislative history, there is usually littleargument that some official documentary source accurately recorded what was stated. See Ross & Tranen, supra note

30

preferences, not claiming a common legislative intent or purpose that would have been consented toregardless of the actual legislative procedure used. While these estimates may sometimes be uncertain,even then they should improve the satisfaction of enactable preferences as long as the estimates are moreaccurate than random guesses.

Still, it may often be the case that enactable preferences do not admit of any reliable estimate. Eventhat does not, however, mean that the default rule approach should be abandoned. To the contrary, thejudicial inability to estimate enactable preferences itself justifies either, as discuss in another paper, apreference-eliciting default rule to obtain an accurate answer to the question about where enactablepreferences lie,93 or a supplemental default rule based on deference to political subunits or judiciallydeveloped views.94 The same applies, only more broadly, for those who have a more thorough-goingskepticism of any judicial ability to estimate of legislative preferences. Judge Easterbrook, for example,has written that agenda manipulation and logrolling have such extensive effects on legislative outcomes that“judicial predictions of how the legislature would have decided issues it did not in fact decide are boundto be little more than wild guesses.”95 This strikes me as somewhat extreme, in part because theinterpreting judge need not guess what precisely the legislature would have done, but which among somelimited set of interpretive options the legislature is more likely to have chosen. But even if one sharedEasterbrook’s bleak assessment, it would not mean default rules to maximize political satisfaction have noplace. It would merely mean that one should favor more extensive application of preference-eliciting orsupplemental default rules.

Objection 2. Legislative History Is Inaccurate. Another argument against using legislativehistory is that, even if a coherent legislative intent existed, the legislative history would not accurately revealit. Commentators from Max Radin to Justice Scalia have argued statements by legislators or legislativecommittees offer a very inaccurate picture of legislative intent.96 They stress that legislators or committeesmay make statements strategically, to lull opponents or mislead interpreters, even if the speaker himselfdoes not believe them. Further, even if the legislative statements accurately reflect the speakers’ ownpreferences, we cannot assume those preferences were read or shared by other legislators. Althoughnormally couched as arguments regarding legislative intent, these points seem equally valid in analyzing theuse of legislative history to estimate enactable political preferences.

Others have disputed this claim of great inaccuracy in legislative statements, or (perhaps more tothe point) denied it is less accurate than the alternative methods of resolving statutory ambiguities.97 They

, at 224-25. The dispute is just whether what was stated reflects correctly the legislature’s views. 98 See, e.g., Farber & Frickey, supra note , 74 VA. L. REV. at 445.99 McNollgast, Legislative Intent, 57 Law & Contemporary Problems 3, 24-26 (1994) [hereinafter McNollgast,

Legislative Intent]; McNollgast, Positive Canons, supra note , at 726.100 McNollgast, Positive Canons, supra note , at 721.101 Id. at 711-13, 731.102 Cf. Elhauge, Preference-Eliciting, supra note .

31

note, for example, that studies indicate legislators are more likely to read committee reports than statutorytext.98 Moreover, the alternative to relying on legislative history to resolve textual ambiguities is relying ondictionaries, treatises and caselaw, and legislators are even less likely to have read those. Some publicchoice scholars also argue that committee chairs and floor managers act as agents for the legislative majorityand are subject to removal or reputational sanctions if they are inaccurate.99

I do not propose to resolve this longstanding debate here. Instead, I wish to stress how theseproblems can be minimized if instead the question is whether to use a general default rule of inquiring intoall forms of legislative history that bear on enactable preferences. The essential problem with relying onstatements by legislators and committees is that they may be insincere or fail to reflect the views of otherlegislative participants. But a more general enactable preferences default rule would, as noted above, oftenfocus instead on the inferences that can be drawn from other actual legislative enactments or widespreadpolitical clout, each of which is objectively verifiable and harder to fake or hide.

As for statements by legislators and committees, a default rule approach could help in two ways.First, courts could adopt default rules about which forms of legislative history to rely on, and how muchreliance to give them, that are more firmly linked to determining which political views of actually enactable.For example, legislative statements provided by those who oppose a bill and were thus not part of thecoalition necessary to enactment cannot determine which political preferences were actually enactable.100

Likewise, while sponsor and committee statements are invaluable when accepted by other legislators, theyare also likely to be at the extreme end of enthusiasm for the statute and thus should not be regarded ascontrolling when rejected by the median legislators whose agreement was necessary for enactment.101

Second, to deal with the residual uncertainty created by concerns about the possible inaccuracyof any legislative statements that make it through these screening rules, let us divide the issue into twopossible cases: (1) the case where the examined legislative statements accurately reflect enactablepreferences, and (2) the case where they do not.

If accurate, then there is plainly a strong preference-estimating argument for courts relying onlegislative statements. Not because they reveal the actual subjective intent of the government, but becausethey provide the best guess as to the political preferences that would have determined how the governmentwould have resolved the question if it had thought about it. The enacting government would want suchlegislative statements used because the introduction of reliable evidence on its political preferences can onlyincrease the extent to which those preferences are accurately reflected in statutory results.

If inaccurate, then there is a strong preference-eliciting argument for relying on such legislativestatements in statutory interpretation.102 Such interpretive use discourages legislators from strategically

103 Ross, supra note , at 575-76.104See Manning, Textualism as Non-Delegation, supra note , at 686 (collecting sources on such practices). 105 See Jonathan R. Siegel, The Use of Legislative History in a System of Separated Powers, 53 VAND. L. REV.

1457, 1524-27 (2000); McNollgast, Positive Canons, supra note , at 726-27..

32

making statements that do not accurately reflect their own political preferences in order to lull opponents.Knowing that their statements will guide interpretation, legislators will be encouraged either to makeaccurate statements about their own preferences, or (in what amounts to the same thing) to accurately statewhat they are willing to live with on the particular issue to procure the more general statutory result. Suchinterpretive use also encourages opponents to contradict a legislator who makes an assertion about thelegislature’s political preferences that does not match their own. Those who do not voice such objectionswill suffer the penalty of having those adverse statements guide interpretation. Thus, a default rule of relyingon legislative history will encourage both sides to state their disagreement with whatever their colleaguesare saying so that any interpretative problem is out in the open. This will not only make the final set oflegislative statements a more accurate reflection of overall governmental preferences, but will also reduceinaccuracy by alerting the legislature about issues that it may want to resolve explicitly. Thus, even if manylegislative statements are inaccurate, the government should generally prefer a default rule of relying on thembecause it will overall increase the satisfaction of its political preferences.

A default rule of examining legislative history also usefully elicits legislative-wide reactions toimprove the accuracy of its legislative history. If the current forms of legislative history the legislaturegenerates are inaccurate, then under this default rule that will increase the legislature’s politicaldissatisfaction with statutory interpretations. This will give legislatures incentives to take steps to improvethe accuracy of its legislative history. For example, they might publish committee mark ups, identify thebill sponsor on the record, state whether certain articulated views represent a consensus among the enactingcoaltion, and require that committee members sign the committee reports to make sure they reflect the fullcommittee’s views.103 Judges could move that process along by, for example, recognizing that committeereports that are signed by only the chair may not signal majority support at the committee level, and thusare far less determinative of what is enactable.

The above analysis also suggests limits on the reliance on legislative history. Some legislativestatements might not be disseminated widely enough to alert others to respond or might come too late inthe legislative process to allow time for contradiction. This is generally not a problem for the sort ofcommittee reports and sponsor statements normally treated as most authoritative by the court. But it is aserious problem for statements made in final House-Senate conference reports about statutes that are thenenacted with no or truncated debate. Even worse problems are presented by House Reports orstatements delivered too late to be considered by the other House, or statements inserted in theCongressional Record that were never actually stated in Congress,104 or by Presidential Signing Statementsthat are made after all legislative action has ended.105

However, a willingness to refer to legislative history need not dictate a blind willingness to acceptanything that can go under that label. Instead, an interpreter should properly be skeptical of any legislativestatements that were offered under circumstances that did not permit others to contradict them. This is

106 See, e.g., Clarke v. Securities Industrial Ass’n, 479 U.S. 388, 407 (1987); Board of Governors v. DimensionFinancial, 474 U.S. 361, 372 (1986); Monterey Coal Co. v. Federal Mine Safety & Health Rev. Comm'n, 743 F.2d 589 (7thCir. 1984); Costello, supra note , at 43-59 (collecting cases).

107 Ross & Tranen, supra note , at 235.108 See supra at __.109 See, e.g., Civil Rights Act of 1991, § 105(b) Pub. L. No. 102-166, 105 Stat. 1071 (1991) (“No statements other

than the interpretive memorandum . . . shall be considered legislative history”).

33

clearest if those late minute statements are likely inaccurate. But even if they are likely to be accurate,courts might reasonably ignore them, effectively using a form of preference-eliciting default rule thatencourages legislators to either make such statements earlier in the process, to allow more time forcontradiction, or to put the point into explicit legislative language. Such a default rule would likely produceboth more preference satisfaction and better deliberation. Consistent with this, courts often reject legislativehistory that is placed on the congressional record but that other legislators did not have an opportunity tocorrect before enactment.106

One might fear Presidential statements would always be excluded under such a rule since the billis only presented to the President after it is passed by Congress. Such an exclusion would be distortingsince Presidential views often strongly influence what is actually enactable. But the President can alwayshave his legislative allies put his views on the Congressional Record.107 If the President does not haveenough legislative support to do even that, he won’t have the one-third support he needs to sustain a vetoand make his views relevant to determining what preferences are enactable.

Finally, an important implication follows from the premise that the whole preference-estimatingapproach of examining legislative history is just a default rule. That premise means the legislature canalways opt out of this default rule if the legislature determines the legislative history on balance will decreasethe accuracy with which its preferences are estimated. In actual practice, as we have seen, thoselegislatures to consider the question have done precisely the opposite.108 This strongly suggest thatlegislatures themselves do not believe that judicial reliance on legislative history reduces the satisfaction oflegislative preferences.

True, even if this is a generally sensible default rule, it might not be advisable for particular formsof legislative history or particular statutes. But legislatures can tailor opt outs for particular statutes or formsof legislative history, and have done so at least once.109 The availability of this rarely used failsafe does notundermine, but rather confirms the wisdom of the general rule, because it means legislatures can deviatefrom the general rule when merited by particular circumstances.

Objection 3. Using Legislative History Is Unconstitutional. The objection might still remainthat, whether or not using legislative history improves the extent to which statutory results accurately reflectlegislative preferences, its use is unconstitutional.

The standard argument is that courts should not rely on legislative history because it has not satisfiedthe constitutional requirements for enacting statutes – in the United States federal government, bicameral

110 See, e.g., Thompson v. Thompson, 484 U.S. 174, 191-92 (1988) (Scalia, J., dissenting) (“Committee reports,floor speeches, and even colloquies between Congressmen, are frail substitutes for bicameral vote upon the text of a lawand its presentment to the President.”); John F. Manning, Constitutional Structure and Statutory Formalism, 66 U. CHI.L. REV. 685, 687 & n.12 (1999) (collecting sources); Manning, Textualism as Non-Delegation, supra note , at 696-99(same); Easterbrook, Text, supra note , at 68-69; cf. Easterbrook, Statutes’ Domains, supra note , at 537-39 (suggestingsimilar argument).

111 The points in this paragraph are discussed in greater detail in Elhauge, Preference-Eliciting, supra note , atXI.C.5, which addresses the general textualist case for an alternative default rule of narrow construction.

112 See Bank One Chi., N.A. v. Midwest Bank & Trust Co., 516 U.S. 264, 279-80 (1996) (Scalia, J., concurring inpart and concurring in the judgment); Eskridge & Ferejohn, Game, supra note , at 552 (describing this as one of Scalia’sargument); Manning, Textualism as Non-Delegation, supra note , at 675-76, 706-37.

113 Manning, Textualism as Non-Delegation, supra note , at 715-19 (collecting cases barring legislativedelegations to one house of the legislature, to a subset of legislators, or to an official under the control or removable bythe legislature, and arguing that they require rejecting legislative history).

114 Manning, Textualism as Non-Delegation, supra note , at 698 (describing this as one of Scalia’s arguments).115 See id. at 727-28.116 Id. at 699-705.

34

adoption by both legislative houses and presentment to the President.110 Courts should instead rely onlyon the statutory text that did comply with these constitutional requirements. But because no text can beinterpreted without some resort to extrinsic evidence, one cannot really say that interpretive aids can belimited to the text.111 Further, the constitution simply leaves statutory interpretation to be resolved by the“judicial power” without specifying how that interpretation should be conducted, and to the extent we haveevidence on the original understanding of this phrase, it appears to have conferred broad power on courtsto go beyond the text in interpreting statutes when that helps courts advance legislative preferences. Finally,the proposition that courts should encourage clarification through statutory text, unless imbedded in adeeper theory, lacks any logical stopping point, suggesting the adoption of the worst possible interpretationin all cases.

The more serious question with extrinsic evidence is not whether to rely on it, but what types ofextrinsic evidence we should use. Thus a more clever argument against the use of legislative history, offeredby Justice Scalia and well-developed by Professor Manning, is that the problem is not with all uses ofextrinsic evidence, but rather with relying on statements by legislative subgroups.112 They point out thatlegislative committees are agencies controlled by Congress. Thus, they argue, relying on their statementsas authoritative interpretations amounts to an allowing an improper delegation of administrative powers toan agency controlled by Congress, which violates constitutional separation of powers principles.113 Thisargument should be distinguished from the argument, also made by Scalia and others, that relying onlegislative history violates the general non-delegation doctrine that limits the legislature’s ability to delegateits legislative powers to anyone else.114 The general non-delegation doctrine has little if any bite,115 and inany event would not explain why courts should reject legislative history but rely on dictionary or treatisewriters, agencies, or judicial precedent.116 Instead, the more telling argument is that relying on legislativehistory allows legislatures to violate the separation of powers prohibition against legislative self-delegation.This argument rests on an objection to legislative self-aggrandizement, rather than to legislative self-

117 The argument thus has less weight in systems less concerned about maintaining a balance of powersbetween legislatures, executives, and courts.

118 See Manning, Textualism as Non-Delegation, supra note , at 675 n.9.119 Id. at 725.120 See supra at __; infra at VI-VII.121 See Manning, Textualism as Non-Delegation, supra note , at 679-80, 683-84, 721.122 Further, as noted above, such statements should be treated as much more authoritative when they reject a

certain interpretation (assuming the committee has de facto veto power) than when they embrace a certain interpretation.123 See US CONST, ART I, § 5, cl 2.

35

abnegation.117

There are several responses to this interesting constitutional theory. To begin with, even if valid, thistheory is not inconsistent with the default rule approach emphasized here. Professor Manning himselfconcedes that this anti-self-delegation principle does not bar courts from examining legislative history thattakes the form of other actual legislative acts or failures to act – like the adoption or rejection ofamendments or the enactment or failure to enact other statutes.118 And of course it provides no bar torelying general political forces or on agency interpretations.119 Both those form major parts of a sensibledoctrine of preference-estimating default rules.120

His theory might seem more directly in conflict with respect to his treatment of statements bylegislative committees and sponsors. But what proves to be the crucial point in this argument is that somecases are treating the legislative committee and sponsor statements as “authoritative” rather than just as onepiece of possibly relevant extrinsic evidence.121 This is what justifies the claim that relying on suchstatements allows a self-delegation of actual administrative power. Under a preference-estimating defaultrule, in contrast, such legislative statements would be only one relevant piece of evidence, albeit animportant one, within a larger set of possibly more important evidence for inferring what politicalpreferences were enactable.122 Indeed, a major advantage committee and sponsor statements havecompared to other sources of extrinsic evidence (like dictionaries, treatises, and caselaw) is that they aremore likely to alert other legislators who are necessary for the enactment to object when those statementsdo not accurately reflect their views, an opportunity for objection that can only be meaningful if thecommittee or sponsor statements are not treated as irrebuttable. Professor Manning’s argument, ifcredited, would thus explain why such legislative statements should not be given authoritative weight, butnot why they should be ignored altogether.

The truth is that judges do not rely on committee and sponsor statements because they have anyinherent authoritaty. Judges rely on them only because (and to the extent) as a matter of legislative practiceother legislators rely on such statements to explain the import of the bills on which they are voting. Andeach legislative chamber has direct authority under the Constitution to determine its own rules forconsidering bills.123 To refuse to recognize such actual legislative reliance not because it is irrelevant butbecause a court considers the reliance illegitimate might even violate the Constitutional provisions vestingthat rulemaking authority solely in the respective chambers. In any event ignoring such committee andsponsors statements would certainly be unfortunate since, especially when other legislators acquiesce in

124 See Manning, Textualism as Non-Delegation, supra note , at 706-07.125 See Elhauge, Preference-Eliciting, supra note , at XI.C. Moreover, even when the conditions are met, judicial

interpretations that ignore legislative history may not put the burden of eliciting legislative reaction on the right side ofthe political controversy and thus fail to elicit a legislative reaction at all.

126 See id.127 See supra at __.128 See, e.g, CLARK, CORPORATE LAW 29, 219-221, 234-262, 761-800 (1986)..129 Sometimes corporations change their corporate identity when they go public. But they do not always, and

in any case even when they do, they rarely define matters covered by default rules, like the proper scope of fiduciaryduties such as the corporate opportunity doctrine.

36

them, they constitute the best evidence of what political preferences were actually enactable. Indeed, Professor Manning himself ultimately argues that judicial interpretations that do not use

legislative history are desirable precisely because they are more likely to deviate from legislative preferencesand thus provide a structural incentive for Congress to clarify more in the statutory text.124 This argumentimplicitly concedes that legislative history does help accurately estimate legislative preferences, but invokesa preference-eliciting argument for thwarting legislative preferences to provoke legislative clarification.Although a preference-eliciting default rule is sometimes desirable, the conditions when this is true arelimited, and when they do not hold imposing a preference-eliciting default rule systematically increasespolitical dissatisfaction.125 Moreover, without those limiting conditions, the principle of encouraginglegislative specification lacks any logical stopping point, but rather would generally require the worstinterpretation the court can imagine.126 In terms of preventing self-delegation, Manning’s argument also hasthe internal problem that legislatures usually also delegate the drafting of statutory text to committees andsponsors, and that other legislators review the text even less closely than they review the committee reportsand sponsor statements.127

C. Updating for Changed Circumstances, Opting Out, and How Hermeneutic Inquiry Differs

Static preferences need not dictate a static default rule. Consider the following example fromcorporate law. Even if one assumes a static preference for the most efficient default rule, it is well acceptedthat the corporate default rules used for closely held corporations often do and should differ from thoseused for publicly held corporations.128 But if this is correct, it follows that the proper corporate default ruleswill often change as the corporation grows from closely held to publicly held. Thus, if one asked thequestion which default rule the corporation’s original shareholders would have preferred at the time ofincorporation, the answer is not simply the default rule that is best for closely held corporations. For sucha static default rule will, when the corporation grows, produce inefficient results, the prospect of which willin turn depress the original share price. Instead, the original shareholders would prefer a dynamic defaultrule that starts out as the rule best for closely held corporations but changes to the rule best for publicly heldcorporations as the corporation grew. That will maximize the original share price, which in part impoundsthe likely future price if the corporation goes public.129

Take corporate opportunities doctrine. Few corporate charters define which business opportunities

130 See CLARK, supra note , at 234-262 (collecting cases that fit this account).131 Id.132 See DWORKIN, supra note , at 348-50; ESKRIDGE, DYNAMIC, supra note , at 5-6, 9-11, 48-55, 107-08, 111-12, 120-

21, 124-30; RICHARD POSNER, OVERCOMING LAW 231 (Harvard University Press 1995); GUIDO CALABRESI, A COMMON LAW

FOR THE AGE OF STATUTES 2, 6-7, 82, 100-04 (Harvard University Press 1982); G. GILMORE, THE AGES OF AMERICAN LAW

95-98 (1977); Lawrence Lessig, Fidelity in Translation, 71 TEX. L. REV. 1165 (1993); Eskridge, Overriding, supra note, at 390-91; Sunstein, Interpreting Statutes, supra note , at 405; Daniel A. Farber, Statutory Interpretation andLegislative Supremacy, 78 GEO. L.J. 281,282-83, 306-17 (1989); T. Alexander Aleinikoff, Updating StatutoryInterpretation, 87 MICH. L. REV. 20, 21-22, 46-61 (1988).

37

are corporate opportunities unavailable to corporate fiduciaries. We thus need some default rule. Supposethe following standard analysis is correct.130 Because fiduciaries in closely held corporations often haveoutside business interests in the same line of business as the corporation, every business opportunity thatlies within the corporation’s line of business should not be considered a corporate opportunity, or else therule would discourage skilled persons from participating in close corporations. But in publicly heldcorporations, where officers generally work full time for the corporation, it makes more sense to defineeverything within the corporation’s line of business as a corporate opportunity. If this analysis is true, thenthe appropriate corporate opportunity default rule changes as the corporation grows from closely held topublicly held. And the original shareholders would prefer a default rule that adjusts over time to reflect this.This appears to be implicitly accepted by corporate law, which does in fact vary the corporateopportunities doctrine for closely held and publicly held corporations.131

This is not to say that the courts would be justified in changing an explicit charter provision oncorporate opportunities. If the corporate founders wished to opt out of judicially set default rules entirelyby providing an explicit provision, their preference for that explicit term would be thwarted by judicialmodification. If the corporate founders prefer dynamic modification, they are free to leave gaps orambiguities that effectively delegate the matter to ongoing judicial resolution. But where they have enactedan explicit term, it must be given effect until the charter itself is amended.

In some applications, giving effect to explicit terms will no doubt turn out to be unwise andundesirable to the corporate founders. But by opting out of judicial default rules, the founders haveindicated that they view these costs of having an inflexible explicit term as lower than the costs of judicialerror in filling and dynamically altering a default. The corporate founders are entitled to trade off the greaterflexibility of judicial ex post modification against the greater accuracy from tailoring a fixed rule themselves.

The above sort of rationale for -- and limits on -- dynamic default rules in corporate law will alsooften be true for those who enact legislative statutes. Pursuing the enactors’ preferences may well call fora default rule that responds to changes in technology, society, the economy, and surrounding legalframework. Courts are then justified, in cases of statutory ambiguity, in updating the default rule along withthose changed circumstances. Enacting governments may always opt out of this dynamic default rule, butunless they opt out should be understood to have accepted updating of the default rules used to resolve theambiguities or gaps they left in statutory meaning.

This provides a more solid (but more limited) ground for the proposal by many notable scholarsthat courts should update and even overrule “outdated” statutes to adjust for changed circumstances.132

38

Where statutory meaning is ambiguous, such a dynamic default rule to fill the statutory gap is appropriate.But if the statute is taken to have a fixed meaning, then the government that enacted that meaning haseffectively opted out of any default rule, including dynamic default rules. The inflexibility costs of adoptingsuch a fixed meaning would surely be significant when circumstances change. But it would be entirelyrational for an enacting government to conclude that (for some issues or statutes) such inflexibility costswould be outweighed by judicial errors if courts were freely permitted to update statutory meanings toadjust for changed circumstances. As honest interpretative agents, courts cannot decline to follow sucha legislative opt-out, and thus are not free to deviate from any fixed meaning they find in the statute.

But while bound by what the enactors actually meant, courts need not be bound by hermeneuticinquiries into what the enactors most likely thought the statutory language meant. Asking that question (incases where meaning is ambiguous) is likely to produce a meaning restricted to expectations that wereconstrained by the circumstances that existed at the time of enactment. And a meaning fixed by thosecircumstances will likely thwart the enactors’ preferences when circumstances change.

Again corporate law offers a fruitful analogy. If some newsreporter asked a founding shareholderin a closely held corporation what specific fiduciary duties he means to imply by the provisions creating thecorporate office of Treasurer, he is likely to provide those specifics that make the most sense for a closelyheld corporation, including a freedom to pursue other business opportunities. But if he has not specifiedthis as a fixed meaning in the charter, he should not be understood to have opted out of the normal defaultrule that fiduciary duties change as the corporation grows. To do so would thwart his likely interest inhaving the most efficient duties, without any indication that he intended that result.

This analysis provides my own theory for the proper dividing line between default rule inquiries intogovernmental preferences and hermeneutic inquiries into statutory meaning. The inquiries share similarities,but in my analysis the line is crossed when the enacting government creates indicia that it desires to opt outof any process of judicial gap-filling or updating. Such indicia should be understood to bear on statutorymeaning and not just preferences, because when persuasive they indicate a legislative opt-out from defaultrules. But if (like the shareholder in the close corporation above) the enacting government leaves no indiciathat it meant to opt out of normal dynamic default rules, then the relevant evidence should be taken toestablish not the most likely fixed meaning, but rather the most likely political preferences out of which toform a default rule that will operate on changed circumstances.

I hasten to add that the remainder of the analysis stands whether or not one accepts this definitionof the limits to hermeneutic inquiry. Wherever you define the limits on that inquiry, you need a set of defaultrules that take up where hermeneutic inquiry leaves off. And the foregoing means that those default rulesshould be updated for changed circumstances whether or not you believe that statutory meaning (or otherhermeneutic conclusions) should be updated as well.

IV. WHEN NO ONE OPTION IS MORE THAN LIKELY TO REFLECT ENACTABLE PREFERENCES

Most rational choice models assume that judges pick statutory interpretations from an infinite set

133 See ESKRIDGE, DYNAMIC, supra note, at 167-70; William N. Eskridge & John Ferejohn, Making the Deal Stick ,8 J. L. ECON. & ORG. 165, ___ (1992); William N. Eskridge, Jr. & John Ferejohn, The Article I, Section 7 Game, 80 GEO.L. J. 523, 549-51 (1992); Ferejohn & Weingast, supra note , at 267-68; Bill Eskridge, Overriding Supreme Court StatutoryInterpretation Decisions, 101 YALE L.J.. 331, 378-85 (1991); Schwartz, Spiller and Urbiztondo, A Positive Theory ofLegislative Intent, 57 LAW & CONTEMP. PROBS. 51, 56-57 & n.23 (1994). See also Linda Cohen & Matthew Spitzer,Solving the Chevron Puzzle, 57 LAW & CONTEMP. PROBS. 65, 69-76 (Spring 1994) (assuming there is a best statutoryinterpretation but that courts are free to deviate from it across an infinite continuum of views about policy and process);Gely & Spiller, A Rational Choice Theory of Supreme Court Statutory Decisions with Applications to the State Farmand Grove City Cases, 6 J.L. ECON. & ORG. 263, 268-83 (1990) (basing most of analysis on assumption of infin i tecontinuum but also considering possibility court can only make a yes-or-no choice). A slightly different take is offeredin Pablo T. Spiller, Rationality, Decision Rules, and Collegial Courts, 12 INT. REV. L. & ECON. 186, 188-90 (1992), whichassumes judges always have interpretive discretion to choose among any point in a policy space, and would generallyprefer to do so, but are sometimes driven by cycling problems to instead decide on a yes-or-no basis in order to achievestable agreement on the point most likely to come close to the judicial policy preferences. None of these articlesconsiders the possibility that hermeneutic methodology might reduce the interpretive choices to, say, three possibilities.

134 Because the probabilities are relative, they will add up to 100% even though they might also be positive oddsthat legislative preferences or deadlock would have prevented any of the interpretive options from being enacted. Seesupra note . For example, if the odds of deadlock were 50%, and the odds of 1, 2 and 3 being enacted were respectively20%, 15% and 15%, then the relative odds of enactment if the government had made a choice would respectively be 40%,30% and 30%.

135 See Ian Ayres, Making a Difference: The Contractual Contributions of Easterbrook & Fischel, 59 U. CHIC.L. REV. 1391, 1402 (1992); David Charny , Hypothetical Bargains, 89 M ICH. L. REV. 1815 (1991). In contracts and corporatecharters, the problem arises in a somewhat different guise, because the usual question is what to do when the populationentering into contracts or corporate charters differ in what default rule they would prefer, but the court cannot tell themapart and no single default rule seems likely to command a majority. Thus, for contracts and corporate charters thisquestion would normally be framed as which of the “minority” options to choose as a general default rule where tailoringseems impossible. Here, in contrast, I am assuming courts are tailoring and the percentages do not refer to the share ofthe population that favors a particular statutory interpretation, but rather refer to the estimated odds that theinterpretation could have gotten enacted if the enactors had been required to make a choice. Of course, this questioncan also arise for tailored contract or corporate default rules where the court is tailoring the default rule and is uncertain

39

of options in a policy space.133 This is an unrealistic depiction of actual statutory interpretation. Althoughlegal methods of statutory interpretation often fail to point to one unambiguous interpretation, legalmethodology is sufficiently constraining that it normally produces some limited set of legally plausibleinterpretations, from which the court must choose.

Nonetheless, often there will be more than two legally plausible interpretive options, meaning thatnone of the options might be the one the legislature “most likely” would have chosen if it had made a choice.What should courts do when there are multiple interpretive options, none of which seems more than 50%likely to be what the enactors would have chosen if it had made a choice? Suppose, for example, thereare three plausible interpretive options, and the court’s best estimate of the relative odds of what thegovernment would have enacted had it made a choice is 40% for option 1, 30% for option 2, and 30% foroption 3.134 Should the court just go with option 1 because its odds of matching enactable preferences arehigher than those of the other options?

A similar theoretical problem has been raised, but never satisfactorily resolved, for contract orcorporate default rules.135 But under the general framework of this article, the abstract answer seems clear.

what the parties would have wanted.

40

The goal of an honest interpretive agent is to maximize the extent to which interpretations satisfy thepreferences of the text-making authority. An interpreting court should accordingly choose the interpretiveoption that minimizes the expected preference dissatisfaction of the parties making the legally binding text(be it a contract, charter, or statute). In the case of statutory interpretation, the court should choose theoption that minimizes the expected dissatisfaction of enactable preferences. But the analysis extends inobvious ways to contracts and charters too.

Perhaps surprisingly, this is not the same as choosing the option that has the highest odds of beingright. A diagram might help. Suppose statutory uncertainty leaves us with three possible interpretiveoptions that can be plotted on a line where the distance between options indicates the extent to which thepreferences of a enacting government desiring one option would be dissatisfied with the other option. Let’s

call the options Narrow, Middle, and Broad, with each separated by 10 units of preference dissatisfaction:that is, if the enacting government would have preferred the middle option, it suffers 10 units ofdissatisfaction from either the narrow or broad options; if it would have preferred one of the extremes, itsuffers 10 units of dissatisfaction from the middle option and 20 units from the opposite extreme. Supposefurther that while a court cannot be sure which option best fits enactable preferences, its best estimate isthat the broad option is 40% to match enactable preferences, with the other options each 30% likely.Choosing the broad option is more likely to fully satisfy enactable preferences, but overall will producemore expected preference dissatisfaction than necessary. If the court chooses the broad option, theexpected preference dissatisfaction will be (.4)(0) + (.3)(20) + (.3)(10) = 9. But if the court chooses themiddle option, the expected preference dissatisfaction will only be (.3)(10) + (.3)(0) + (.4)(10) = 7. Thecourt should thus choose the middle option even though it is less likely to match enactable preferences thanthe broad option.

We can generalize the analysis using a little bit of math. Suppose we have three interpretativeoptions (A, B and C) that can each be conceptualized as points along a policy continuum so that the further

136 This follows because if B is some middle point, the distance from A to B, plus the distance from B to C, mustadd up to the distance between A and C. Also, since A and B are by hypothesis different points, DAB > 0.

41

the point is away from the enactors’ ideal point the more dissatisfied they are, and with B signifying somepoint between A and C. Let DAB stand for (the absolute value of) the dissatisfaction resulting from adifference between the options A and B, with DAC and DBC meaning the same for the differences betweenoptions A and C, and B and C. Suppose further that the relative probability that, had the enactors madea choice, they would have enacted A is PA, for B is PB, and for C is PC. Then, the expected politicaldissatisfaction from choosing each point will be:

If choose A: PB DAB + PC DAC

If choose B: PA DAB + PC DBC

If choose C: PA DAC + PB DBC.A judge who is trying to minimize political dissatisfaction will choose the option with the lowest expecteddissatisfaction value. Thus, this judge should choose middle option B if both:

PA DAB + PC DBC < PB DAB + PC DAC ANDPA DAB + PC DBC < PA DAC + PB DBC

I first solve to see when middle option B would be a better choice than end point option A even though Ahas the highest odds of matching enactable preferences. Since PA + PB + PC = 1, we know that PB = 1 -PA - PC. Thus, the judge should choose B over A if:

PA DAB + PC DBC < (1 - PA - PC)DAB + PC DAC, which can be rearranged as:(2PA - 1) DAB < PC (DAC - DBC - DAB).

Since B is some point between A and C, DAC - DBC - DAB necessarily equals 0, no matter where optionB lies between options A and C.136 Thus, the above equation produces the result that the judge shouldchoose option B over A if 2PA - 1 < 0, or PA < .5. For example, option B is the better choice even ifoption A is 49.9% likely to match enactable preferences, and options B and C are 24% and 26.1% likelyto match enactable preferences. Indeed, middle option B is the better choice as long as the probability thatend point option A will match enactable preferences does not equal or exceed 50%.

Where end point A has the highest odds of matching enactable preferences, middle option B is alsoalways a better option than end point option C. Option B will produce less dissatisfaction than option Cif:

PADAB + PCDBC < PADAC + PBDBC

Again, PB = 1 - PA - PC. Thus, the judge should choose B over C if:PADAB + PCDBC < PADAC + (1 - PA - PC)DBC,

which can be rearranged as:(2PC - 1)DBC < PA(DAC - DBC - DAB).

Again, DAC - DBC - DAB necessarily equals zero (and DBC is positive since B and C are not the same point),thus the condition is met whenever 2PC - 1 < 0, or when PC < .5. Since by hypothesis A is the option withthe highest odds of matching enactable preferences, this must always be true.

The way to see intuitively why middle option B is always preferable to end point A is to begin withthe extreme case where PA = 49.9% and PB = 0%. In this case, PC must = 50.1%. A decisionmaker

137 A fortiori, option C cannot be preferable to A because option B dominates C.138 See supra at III.B.

42

would prefer option B over A because while the move from A to B increases dissatisfaction by thedifference between B and A with probability 49.9%, it decreases dissatisfaction by the same difference withprobability 50.1%. Of course, this example does not satisfy the hypothetical assumption that A has thehighest odds of matching enactable preferences. But now suppose we alter this hypothetical to transferfrom option C to option B any percentage necessary to make A the highest odds option without increasingA to over 50% (here any percentage > 0.2%). Such a transfer can only make B a more attractive optionthan it was before (since we are increasing the odds that it precisely matches legislative preferences), andthus B continues to be preferred over A. Morever, since B is preferable to any end point A (whoseprobability is below 50%), it must necessarily be preferable to any end point C (whose probability is lowerthan A).

Can middle point B ever prevail over option A if it is more likely than not that option A matcheslegislative preferences? No. Under this assumption PA > .5. And then it will be impossible to satisfy thecondition that 2PA - 1 < 0.137 The way to see this point intuitively is to take the extreme case where thereare only two points, A and C, with A thus 51% likely to be the correct choice. Obviously, A would be theright choice over C. This cannot be altered by switching some percentage from C to an option B that iscloser to A, for that only makes A more desirable.

The interesting result is that when, among the available interpretive options, no one option ismore likely than not to match enactable preferences, a court trying to minimize expected politicaldissatisfaction should choose a moderate interpretive option even though extreme options are morelikely to match enactable preferences. Note that this proposition goes beyond the point, consideredabove, that in estimating what political preferences were enactable, the court should focus on the marginallegislators whose agreement was necessary to enact the legislation.138 That point is true, and the views ofthat marginal legislator should be followed when a court can determine with greater than 50% odds thattheir preferences would have governed. But suppose that, after going through this thought process, onedetermines that no one interpretive option is more than 50% likely to reflect enactable preferences. Then,even if one predicts the marginal legislator would be more likely to pick an option that, among the availableinterpretations, is more extreme, the court should nonetheless choose a more moderate option.

Of course, courts will rarely have these sorts of precise percentages in mind, but one can employthe same insight in the form of a cautionary axiom. The axiom would simply provide that, where multipleinterpretive options seem plausible, courts are justified in favoring middle ground options when interpretingwhat the legislature would have wanted. One often sees just such a tendency in statutory interpretationcases, in part because the need to get a majority of multi-member appellate panels naturally encouragesit. Indeed, this proposition provides one justification for the whole appellate structure of using multi-judgepanels.

V. THE THEORY BEHIND A CURRENT PREFERENCES DEFAULT RULE

139 See supra at III.C.140 See RONALD DWORKIN, LAW’S EMPIRE 347-50 (Harvard University Press 1986); WILLIAM ESKRIDGE, DYNAMIC

STATUTORY INTERPRETATION 5-7, 9-11, 69, 107-08, 120-21 (Harvard Univ. Press 1994) [hereinafter ESKRIDGE, DYNAMIC];GUIDO CALABRESI, A COMMON LAW FOR THE AGE OF STATUTES 2, 6-7 (Harvard University Press 1982); Cass R. Sunstein,Interpreting Statutes in the Regulatory State, 103 HARV. L. REV. 405, 495 (1989) [hereinafter Sunstein, InterpretingStatutes].

141 See DWORKIN, supra note , at 349 (looking to whether interpretation is “fair” given current “public opinion”);Sunstein, Interpreting Statutes, supra note , at 495 (interpretation may change if statutory provision is “no longerconsistent with widely held social norms”); William N. Eskridge, Jr., Public Values in Statutory Interpretation, 137 U.PA. L. REV. 1007, 1019-61 (1989); ESKRIDGE, DYNAMIC , supra note, at 10-11, 156-59 (looking at modern “socialdevelopments, and current values and social needs” and looking to “majority preferences” even though they would notactually be enactable because the actual legislative system is responsive to an interest group minority). Eskridge doesargue that courts must make sure not to adopt interpretations that will be legislatively overriden, id. at 7, 69, but not thatthey should adopt the interpretation most likely to be enactable. Indeed, he even argues in favor of statutoryinterpretations that is sometimes “counter- or nonmajoritarian.” Id. at 151. Calabresi argues that courts should overruleold statutory meanings that are no longer enactable by the current legislature, CALABRE S I, supra note , at 2, 6-7, butwould not require any evidence that the new meaning is the one most likely to be enactable.

142 Dworkin argues that judge should act as legislature’s “partner” (rather than agent) developing the statutoryscheme in “the best way,” which will turn in part on the judges “own judgment,” and rejects the view that “statutesshould be read, not according to what judges believe would make them best, but according to what the legislators whoactually adopted them intended, and arguing that interpretive questions “must be answered in political theory, by takingup particular views about controversial issues of political morality.” DWORKIN, supra note , at 313-14, 316. See alsoid. at 321 (judge’s “judgments will also be sensitive to his convictions about the relative importance of fairness”); id.at 334 (“A judge must ultimately rely on his own opinions” and “[h]is own political convictions”); id. at 347(interpretation should turn in part on the judge’s “substantive opinion” about which statutory reading “would make itbetter from the point of sound policy”); id. at 354 (which cases are easy or hard turns on the judge’s personal“convictions about justice and fairness”). Eskridge likewise regards it as unavoidable that courts must make “valuechoices,” and ultimately supports his theory of dynamic interpretation with the argument that it is “normativelydesirable” for courts to adopt a stance of critical pragmatism that considers whether an interpretation will “unsettle

43

When statutory meaning is unclear, how should courts fill in the default rule when the preferencesof the current government differ from those of the enacting government? This problem is different from thecase of changed circumstances, for in that case one could readily conclude that changing interpretationswould better satisfy the enacting government’s political preferences.139 Here, it may be the case that thecircumstances are largely the same, and that any change would have to be justified by changed politicalpreferences about how to respond to those circumstances.

Some of the most eminent legal scholars, including such luminaries as Ronald Dworkin, BillEskridge, Guido Calabresi, and Cass Sunstein, have taken the position that statutory interpretation shouldbe sensitive to changes in political preferences.140 But their analysis differs from mine in at least fourimportant respects that are important to emphasize at the outset.

First, these scholars would have courts look to loose understandings about current public opinionor social norms, whether or not there is any evidence those reflect the preferences most likely to beenactable in the current legislature.141 Such an approach requires the exercise of substantive judicialjudgment about which of the unenactable current public opinions or values are meritorious enough to governstatutory interpretation.142 Indeed, these scholars would make statutory interpretation turn on a judicial

existing practice,” “reasonably accommodate a diversity of interests,” and be “responsive to social needs and publicvalues.” ESKRIDGE, DYNAMIC, supra note, at 174-76. See also Sunstein, Interpreting Statutes, supra note , at 412-13(rejecting view that “controversial views about public policy . . . should never be part of statutory construction” andsetting forth the “substantive norms” courts should follow and those it should reject).

143 See ESKRIDGE, DYNAMIC , supra note, at 157-61; DWORKIN, supra note , at 319 (interpretation should reflectjudicial “opinions about the influence that attitudes, belief, and ambitions of particular groups of officials and citizensought to have in the process of legislation”) The most common argument is that statutory interpretation should leanagainst special interest groups because they have disproportionate political influence. ESKRIDGE, DYNAMIC, supra note,at 157-58; DWORKIN, supra note , at 319 (judicial interpretation should take into judge’s views on “whether lobbying,logrolling, and political action committees are a corruption of the democratic process”). But as I showed in a prior article,any claim that a special interest group has disproportionate influence depends on an underlying normative judgmentabout the degree of influence that is proportionate to the group’s justifiable interest, and produces results no differentfrom applying that normative judgment directly. See Elhauge, Interest Group Theory, at __; see also supra at II (rejectingalternative of having interpretation turn on substantive judicial judgments, including the judgment that the currentpolitical system underweighs certain interests).

144DWORKIN, supra note , at 312 (judicial interpretation will depend on judge’s “best answer to politicalquestions like “how far Congress should defer to public opinion in matters of this sort”); id. at 319 (judicial interpretationturns on judge’s “views on the old question whether representative legislators should be guided by their own opinionsand convictions” rather than public opinion).

145 Id. at 341.146 Id. at 340-41.147 Id. at 342. See also id. at 349 (same).148 ESKRIDGE, DYNAMIC, supra note, at 151 (emphasis added).149 Id. at 156-57, 159-61.

44

view about which public opinions or majority preferences would be enactable but for certain politicalrealities that these scholars deem normatively undesirable even if they are not unconstitutional.143 Thus,even when these scholars would have judges look to current political preferences, their approach has morein common with theories that advocate the exercise of substantive judicial judgment to resolve statutoryambiguity. They accordingly differ sharply from my approach of adopting statutory default rules thatmaximize the satisfaction of those political preferences that are enactable.

Second, these scholars would not make even this loose understanding of current public opinionbinding on interpretive questions. Dworkin argues that judges must also make their own normativejudgment about whether the issue is an appropriate one for legislators to disagree with public opinion,144

concluding that judicial interpretation should be sensitive to “general public opinion” only when the judge“believes” the statutory issue does not “involve any question of principle” that merits deviation from publicopinion.145 The substantive contestability of which issues involve such questions of principle is neatlyillustrated by Dworkin’s assertion that no such principle is implicated by the preservation of a species.146

In the end, Dworkin only argues that the judge should “not wholly ignore the public’s opinion” becausesometimes the judge’s own view of “political fairness” will make public opinion normatively relevant.147

Likewise, Eskridge argues that “statutory interpretations should often be counter- or nonmajoritarian.”148

In particular, statutory interpretation should in his view favor a minority group if it has been marginalized.149

But, as Professor Tribe showed long ago, judges cannot determine when a minority has too little political

150 See Tribe, The Puzzling Persistence of Process-Based Constitutional Theories , 89 YALE L.J. 1063, 1072-76(1980).

151 See supra at I-II.152 See DWORKIN, supra note , at 347-54; ESKRIDGE, DYNAMIC, supra note, at 5-7, 9-11, 107-08, 120-21; CALABRESI,

supra note , at 2, 6-7; Sunstein, Interpreting Statutes, supra note , at 412, 495.153 See infra at V.E.154 See DWORKIN, supra note , at 347-54; ESKRIDGE, DYNAMIC, supra note, at 5-7, 9-11, 107-08, 120-21; CALABRESI,

supra note , at 2, 6-7.

45

influence without making substantive judgments about the degree of political influence that minority oughtto have.150 And, especially outside the limits of constitutional law, it is unclear what would authorize judgesto impose their substantive judgments on a democratic political process. In contrast, I argue that, wheninterpreting statutes, judges should not exercise substantive judicial judgment but rather should (and do)try to maximize political satisfaction.151

Third, the position of these scholars is not limited to resolving how courts should fill in the defaultrules that resolve ambiguities in statutory meaning. Rather, these scholars argue that current public opinionor values should prevail over enacting preferences even when this requires a change in statutory meaning.152

This view has a fundamental problem that I address in more detail below: the current polity, unlike theenacting one, did not actually complete the process for enacting a text with statutory meaning, and thus hasless claim to authority and less accurately ascertainable preferences.153

Fourth, the underlying rationale of these other scholars is significantly different. These scholarsessentially argue that current public opinion and values should (at least when judges deem themsubstantively important) normatively take precedence when they conflict with enacting legislativepreferences.154 In contrast, my argument will be based on an analysis about what statutory default rule theenacting legislature itself would prefer.

Notwithstanding these fundamental differences, my analysis produces a similar conclusion in favorof dynamic interpretation, albeit one with a more solid basis, more constraining methodology, and morelimited scope. Namely, I conclude that, where there is ambiguity in statutory meaning, the enactinggovernment’s preferences would overall be maximized by a general default rule that dynamically tracks theenactable preferences of the current government – where those preferences can be determined with relativereliability -- rather than statically sticking with the enacting government’s preferences.

This probably seems counterintuitive. One would naturally think that the enacting government’spolitical preferences would necessarily be best maximized by a default rule that stuck to those preferences.And that would indeed be true if we were asking only which default rule would maximize the enactinggovernment’s preference satisfaction for the statutes it enacted. But here we are asking what generaldefault rule for filling all statutory indeterminancies maximizes the enacting government’s politicalsatisfaction. In choosing between default rules, the enacting government would realize that a rule that stuckonly to enacting preferences would maximize its preference satisfaction in the future over the statutes itenacted. But a rule that tracked current enactable preferences would maximize the enacting government’spreference satisfaction during its time in office over all existing statutes, including those enacted by previous

46

legislatures.For example, an originalist position on default rules may mean the 2000 legislature has the pleasure

of foreseeing that its views on the statutes it enacted will fill in statutory ambiguities when those statutes areinterpreted in 2100. But a current preferences position on default rules would mean the 2000 legislaturewould instead have the pleasure of seeing its views fill in the statutory ambiguities for statutes enacted in1900 but interpreted in 2000, not to mention for all other statutes being interpreted in 2000.

Which would more likely maximize the satisfaction of the political preferences held by the enactinggovernment: (1) a enacting preferences default rule that makes future statutory interpretations correspondto its preferences, or (2) a current preferences default rule that makes present statutory interpretation ofall statutes correspond to its preferences? The answer depends both on the accuracy and strength ofpreference determinations in the present versus the future.

A. Current Preferences More Reliably Ascertainable.Consider first the situation where an interpreting court cannot come to any reliable estimate of what

the enacting government’s enactable preferences would have been for filling in a statutory ambiguity, butcan reliably estimate current enactable preferences. This is not a fanciful case, given that currentpreferences are often better known because contemporaneous, and interpretation often involves issues notthought of in the past, or on which any record of thoughts has been lost. In such situations, even theenacting government’s political preferences would clearly be maximized by a general default rule thattracked current preferences. Such a default rule would not lose the enacting government any real influenceover future statutory interpretations since its preferences cannot reliably be estimated in the future anyway.But the default rule would increase the satisfaction of the enacting government’s preferences regardingstatutory interpretations made about older statutes during the enacting government’s time in power. Giventhis trade off, the enacting government would plainly prefer to gain influence over present interpretationsrather than retain an illusory influence over future interpretations. The answer here depends solely on therelative accuracy of preference-estimation, and is thus independent of whether the enacting government hasstronger preferences regarding future or present events.

Consider next the more prevalent situation where a court can estimate both the enacting and currentpreferences, but can estimate the current preferences more accurately. Again, this improved accuracy isa powerful reason why even the enacting government would prefer the court to track current governmentalpreferences. The enacting government gains more satisfaction from having its preferences accuratelyestimated during its time in office than from the prospect they will be less accurately estimated in the future.This holds unless the enacting government’s preferences about future events are not just stronger than itspreferences about the era which it governs, but stronger by a sufficient margin to outweigh the decreasedaccuracy. But for which era are preferences likely to be stronger? To isolate this issue, let us now considerthe case where a court can ascertain either enacting or current preferences with equal accuracy.

B. Current and Enacting Preferences Equally Ascertainable.Suppose an enacting government’s likelihood of having its preferences accurately ascertained by

interpreting courts are equal at present (concerning interpretations of past statutes) and in the future

155 Political preferences that a polity holds but fails to reflect in its choice of legislators would not be enactablepreferences.

47

(concerning interpretations of statutes it enacts). Given that equal accuracy, whether the government wouldprefer a current or enacting preferences default rule turns on whether the government is, in the general runof cases, likely to have stronger preferences about present interpretations or future ones. Which is seemsmost likely to be stronger?

Generally, governmental preferences will be weaker regarding future events than regarding eventsduring the government’s time in office. In the future, only some of those holding the enacting politicalpreferences will still be around. Many statutory interpretations that track current political preferencesinvolve cases where so many decades have passed since enactment that the entire legislative polity thatdictates governmental preferences has likely changed. In other cases, there exists some overlap. But aslong as a significant percentage of the voting population have turned over, the same polity will no longerexist and the enactable political preferences will have changed. Even without any change in members, thelegislative polity may change its political views (and the political representatives it elects) as time passes,perhaps because voters’ life-situation has changed in ways that mean they benefit from a different statutoryresult.

The enacting polity’s political preferences for a given statutory result are thus likely to be strongerat the time the enacting government sits because those who hold those preferences are those whoexperience that result. Accordingly, the enacting polity would generally be willing to trade off having lessinfluence over future interpretation of their statutes for having more influence over current interpretation ofpast statutes. For example, enactable political preferences that hold sway in 2000 are more likely to bemaximized by statutory results that correspond to those preferences in 2000 than by statutory results thatcorrespond to them in 2100.

Moreover, there are a lot more pre-2000 statutes being interpreted in 2000, than there are statutesenacted in 2000 that would be interpreted in the future. The set of older statutes that will be applied at thetime when the enacting government sits will cover the full range of statutory results that affect the satisfactionof its enactable political preferences. The set of statutes it enacts will be smaller in number and cover somemore narrow range of statutory results in the future that might impact on preference satisfaction. Thus, eventhe enacting polity should (in cases of statutory ambiguity) prefer a general default rule that tracks theenactable preferences of the current polity where reliably ascertainable. Indeed, every polity would prefersuch a default since it would maximize the political power of each polity at the point in time when it exists.In addition, a current preferences default rule makes current law more responsive to current democraticwishes, as well as reducing the legislative time that must be devoted to updating statutes.

It bears emphasis that the precise question is which default rule maximizes the satisfaction ofenactable political preferences, not which default rule a particular set of governmental officials might prefer.Normally, those amount to the same thing since a polity’s enactable political preferences are manifestedin its choice of elected officials.155 I thus generally use the short hand phrase “governmental preferences.” But sometimes they might differ, a difference that proves relevant in considering various objections to thegeneral proposition that the enacting polity would prefer a current preferences default rule.

156 See MAYHEW, CONGRESS: THE ELECTORAL CONNECTION 49-73 (1974) (legislators' desire to maximize reelectionleads them to three main activities: advertising, providing particularized benefits like pork and casework, and position-taking).

157 See Elhauge, Preference-Eliciting, at II.158 See MORRIS FIORINA, RETROSPECTIVE VOTING IN AMERICAN NATIONAL ELECTIONS 20-43 (1981); Cohen &

Spitzer, Term Limits , 80 GEORGETOWN L.J. 477, 487-88 (1992).159 See Farber & Frickey, The Jurisprudence of Public Choice, 65 TEX. L. REV. 873, 890-901 (1987) (summarizing

literature).

48

Objection 1:A Preference to Have Something to Run Against. One might hypothesize thatsome political officials might prefer statutory decisions that run contrary to their current political preferencesbecause that gives them something to run against in the next election. This is an intriguing possibility, butthe hypothesis is probably untrue, and even if true would not justify a different default rule.

To be sure, some political scientists claim that legislators run for re-election by announcing publicpositions, not by achieving actual legislative results.156 However, few statutory interpretations are ofsufficient general interest to make position-taking about them useful in a campaign. After all, the statutorygaps or ambiguities at issue were not sufficiently salient to fill in at the time of enactment, and the statutorydecisions that fill them in generally fly below the legislative radar screen, operating for decades withoutbeing legislatively reconsidered. Indeed, if the statutory issue were the sort that were likely to belegislatively reconsidered, that might well counsel instead for a preference-eliciting default rule.157 Andunless having judges adopt contrary statutory interpretations had a significant positive effect on their re-election rates, elected officials would not prefer such interpretations since those interpretations do, after all,run contrary to their own political preferences.

In fact, other political science literature tends to indicate that such contrary interpretations shouldhave a negative effect on re-election rates because voters vote retrospectively based on their somewhathazy perception about the governmental benefits the voters received in the past rather than based onbenefits the candidates promise to deliver in future.158 Thus, even if voters would not blame elected officialsfor the statutory interpretations that run contrary to current preferences, the officials will be negativelyaffected by the fact the electorate generally has had a worse experience than it otherwise could have duringthe past term.

More generally, empirical studies tend to indicate that actual legislative conduct is influenced byboth the legislator’s ideology and the interests of their constituents.159 To the extent this conduct reflectsthe revealed preferences of legislators, these empirical studies undermine the notion that legislators wouldprefer statutory results that conflict with the preferences of both themselves and their electorate. Legislatorsmay be strategic about re-election, but the ultimate purpose of getting re-elected is to accomplish theirpreferred policy goals.

The hypothesis that legislators would prefer statutory results that contradict their politicalpreferences thus seems empirically dubious. Still, even if dubious, the posited hypothesis may strike someas plausible. If it were empirically accurate, would that mean that courts should instead choose a defaultrule of conflicting with current preferences to give officials political fodder for their next campaign? The

160 See supra at I n.__

49

answer remains ‘no.’ Officials may have strategic preferences for contrary interpretations, but suchpreferences would neither be enactable nor political. As I have defined the term, “enactable politicalpreferences” refers to “the political preferences of the polity that are shared among sufficient electedofficials that they could and would be enacted into law if the issue were on the legislative agenda.”160 Astatutory result that runs contrary to governmental preferences – and is preferred precisely because it isso unpopular that it creates an attractive election target – could hardly be said to be enactable. Such apreference would also not correspond to the political preferences that either the polity or officials haveregarding certain statutory results.

Nor is there any reason why statutory default rules should try to maximize careerist objectives orstrategic private aims that officials may harbor but could not actually enact into law. The polity has nointerest in having judges interpret statutes in ways contrary to the polity’s current enactable preferences justso individual officials can advance their personal interest in retaining office. Nor would officials have anylegitimate interest in having statutory decisions thwart the political preferences of themselves and theirconstituents just to strategically advance their personal career interests. To argue that statutory defaultsshould conform to such careerist objectives would be like rejecting efficient default rules in corporate lawbecause imperfections in the market for public offerings or the process of adopting charter amendmentsmean that the promoter-managers who write the corporate charters prefer inefficient default rules thatexacerbate agency costs to profit themselves at the expense of shareholders.

Although judges should serve as honest agents, the principal to whom judges are ultimatelyresponsible is the polity. To be sure, it is the polity as mediated through a particular choice ofrepresentatives and political processes necessary for statutory enactments. But the fact of that mediationdoes not mean that the strategic private interests of those representatives are what should be maximized,any more than a corporate officer with a corrupt board could be said to have a fiduciary duty to aid theboard in its corruption.

Objection 2: Weaker Preferences Regarding Current Experience than Regarding theFuture Application of What One Enacted. One might relatedly hypothesize that government officialshave more of a personal stake in the statutes they themselves enacted, and thus care more about seeingambiguities in those statutes interpreted to match their preferences in the future than they do about seeingpresent statutory ambiguities interpreted to match the enactable preferences of the current polity thatelected them. Again, I doubt this hypothesis is true, and it would not change the conclusion even if it were.

Choosing a current preferences default rule does not void the statute, and thus would not deprivegovernment officials of any legacy from being associated with the statute in the future. Nor would it denythem the satisfaction of having their preferences followed on any matter they had thought enough about tomake clear in the original statute. The default rule would merely cover interstitial matters that were notclearly resolved in the original statute. On such marginal matters, it is doubtful the officials would havestrong enduring preferences they would prefer to see satisfied in the future more than they would prefer tosee their preferences govern all marginal matters resolved during their time in office. In short, officials mayhave a personal stake in the statutes they enacted, but probably not in the statutory ambiguities they

161 One could try to develop a default doctrine that distinguished between ordinary cases (where the legislativepolity had no esp ecially strong interest in the future interpretation of its enactments) and extraordinary cases (wheresuch an interest is especially strong), and make the default rule track current preferences in the former cases and enactingpreferences in the latter cases. But such a distinction seems dubious. To begin with, it is hard to know what to makeof a strong interest that was not translated into statutory meaning – it seems likely either it was not that strong or viewswere sufficiently divergent that any preferences were not enactable. Moreover, I doubt there are administrable criteriafor distinguishing the ordinary from the extraordinary, and the lack of any such criteria will make it hard for courts toresist deciding based on whether they approve of the preferences that existed at the time of enactment becausepreferences that one shares are naturally going to seem more “extraordinary” than others. Indeed, once one took thistack, there would be no sharp distinction between ordinary and extraordinary, but rather an continuum in the degree ofextraordinariness that, at a different point for each degree, would yield if the statutory interpretation were so far in thefuture that the discount for the lack of effect on the polity would outweigh the degree of extraordinariness in its interest.Such tailoring of the default rule for degree of extraordinariness thus seems far less preferable than choosing a generaldefault rule to govern all such cases.

162 See, e.g., Ferejohn & Weingast, A Positive Theory of Statutory Interpretation, 12 INT 'L REV. L. & ECON. 263,266-67 (1992) (making argument based on that assumption, and collecting similar literature); Matthew D. McCubbins,Roger G. Noll & Barry R. Weingast, Administrative Procedures as Instruments of Political Control, 3 J.L. ECON. & ORG.243 (1987) [hereinafter McNollgast, Administrative Procedures]; Matthew D. McCubbins, Roger G. Noll & Barry R.Weingast, Structure and Process, 75 VA. L. REV. 431 (1989); see generally infra at __ (discussing the different

50

enacted. Even if they did, the fact that officials may have some personal stake or pride of authorship wouldnot justify putting aside the interpretive default rule that would maximize the satisfaction of the polity’senactable preferences.

The enacting polity itself is unlikely to have any special stake in the future resolution of statutoryambiguities it created that is so large that it transcends the polity’s interest in actually experiencing statutoryresults that conform to its preferences. The mere fact that the polity made certain enactments might to somebe taken as proving the polity has a special interest in the topics covered by those enactments. But, moreprecisely, what enactment signals is the polity’s relative interest in changing the pre-existing law on thattopic, not that it has a greater interest in the resulting statutory ambiguities about that topic than it has forother topics. The topic provoking enactments may be far less important to the electorate than anothertopic where the existing (unambiguous) statutory meanings are satisfactory. Further, the existence of thestatutory ambiguity is either unforeseen issue or an intentional failure to decide, neither of which tend tosignal an especially strong interest in the issue. Whatever special interest in changing the law motivatedenactment would likely be resolved by the statutory meaning, or, if unresolved, indicate the sort oflegislative stalemate that signals an absence of enactable preferences. Resolutions of statutory ambiguitiesthat tend not to provoke legislative consideration are by definition of weaker interest – but that weakerinterest is as likely to apply to older statutes as to ones that the polity enacts. In any event, even if thelegislative polity does have a greater interest in the topics it chooses for enactment, that interest necessarilywanes the further away in time, as less and less of the polity is around to experience the statutory results.161

Objection 3: Legislators Prefer Durable Legislation. A final objection focuses on thedurability of legislation. Much rational choice literature assumes that an enacting legislature would want itslegislation to be as durable as possible. But most of this literature depends on the assumption that all thelegislature cares about is whether the future interpretation of the statutes it enacted reflect its preferences.162

assumptions between prior rational choice models and this paper).163 See William Landes & Richard Posner, The Independent Judiciary in an Interest-Group Perspective, 18 J.L.

& ECON. 875, 877-79 (1975); see also Terry Moe, The Politics of Bureaucratic Structure, in CAN THE GOVERNMENT

GOVERN? (eds. Chubb & Peterson 1989) (interest groups demand ex ante protection against ex post political deviations).Bidding in their model should be understood to include not just providing money, but providing any means of politicalsupport useful to those legislators.

164 Landes & Posner, supra note , at 877-79.165 Landes & Posner, supra note , at 880-85.166 See Elhauge, Interest Group Theory, supra note , at 90; Jonathan Macey, Transaction Costs and the

Normative Elements of the Public Choice Model, 74 VA. L. REV. 471, 496-99 (1988).

51

Once one realizes that the question is instead what general default rule to chose for interpreting all statutes,it becomes plain the legislature would care not just about the future implications this choice has forinterpreting the statutes it enacted, but also how the default rule affects how all statutes are interpretedduring the time when the legislature sits. This requires the more nuanced inquiry noted above, rather thanany general assumption of a preference for durability.

Landes and Posner have offered a theory about durability that requires a somewhat morecomplicated response. In their theory, legislators “sell” legislation to the highest bidding interest group, andthe willingness of interest groups to bid would be undermined unless the legislators can offer durablelegislation.163 This theory means both the future and present effects make the enacting legislature wantdurability, because the future lack of durability is what leads to the present decrease in interest group bids,which Landes and Posner posit is all legislators try to maximize. One might thus conclude thatgovernmental officials would prefer a durable default rule that tracks enacting preferences rather than adynamic current preferences default.

But the Landes and Posner theory argues that judges should stick to the original statutory meaningthat reflected the bargain between the legislators and interest groups.164 Here, by definition, the statutorymeaning is ambiguous, and thus a particular resolution of that ambiguity was never bid for, nor resolved bythe original bargaining. Because of that ambiguity, the statutory issue can have no durable resolution untila court interprets it. To the contrary, the prospect of future application of a current preferences default rulewould seem to increase the incentive of interest groups to bid more at the time of enactment to secure adefined favorable statutory meaning. Moreover, even if a current preferences default rule lowered theeffective price that elected officials can charge for new legislation, it would increase the effective price theycan charge for official indications of preferences that would alter the interpretation of older (morenumerous) statutes.

So the premise is dubious that elected officials who cared only about maximizing interest group bidswould prefer an enacting preferences default rule. Indeed, even within the Landes and Posner model, theirconclusion was heavily dependent on contestable empirical assumptions about, among other things, thecosts of securing legislation, effect on durability, and discount rate applied to future benefits.165 It is alsoempirically questionable whether judges have any real motive to help legislators maximize interest groupbids.166

Even if elected officials would prefer an enacting preferences default rule because it maximized

167 See Elhauge, Interest Group Theory, supra note , at 48-66.168 See infra at E, VI.C, VII; Elhauge, Preference-Eliciting, supra note , at XI.B.169 See infra at IX.B.170 See infra at E, VI.C, VII; Elhauge, Preference-Eliciting, supra note , at XI.B.

52

interest group bidding, that would hardly settle the societal question of which default rule we shouldnormatively prefer. I have elsewhere argued that interest group theory does not by itself provide anynormative grounds for judges to try to dampen interest group influence.167 But it is equally true that nothingin interest group theory provides any reason to think judges should go out of their way to increase interestgroup influence or bidding. The goal after all is to maximize not the strategic preferences of legislatures,but the enactable preferences of the polity. And those enactable preferences would generally be maximizedby a general default rule that tracked current enactable preferences when enacting and current preferencescan be estimated with equal accuracy.

On the other hand, political preferences could not be maximized unless statutory interpretations,once arrived at, were sufficiently durable to induce some behavioral reliance. This helps reinforce thelimitation that only a reliable official indication of a change in current enactable preferences suffices tochange an interpretation under a current preferences default rule.168 But these concerns provide no reasonfor a court’s first interpretation of a statutory ambiguity not to track current enactable preferences, nor apersuasive reason to bar a court from ever changing its first interpretation to track reliable officialindications of changed enactable preferences.169

C. Enacting Preferences More Reliably Ascertainable.We have one more case to consider: the situation where the interpreting courts can ascertain

enacting legislative preferences more accurately than current ones. Like its converse, this is hardly a fancifulcase. The enacting government, after all, actually passed a statute on the subject at issue. Coalitions werebuilt around that topic and discussions about it ensued. The current government may be easier to gaugegenerally because contemporaneous with the interpreting court, but may also be less revealing on theprecise area at issue.

This is a important countervailing concern. Even if each legislative polity has stronger preferencesabout its present, it can derive little satisfaction from seeing statutory interpretations conform to mistakenestimates of its preferences. Thus, we would expect to see (and do see) that courts follow currentpreferences only when they can reliably be ascertained, normally because memorialized in some officialaction.170 Concerns about reliability thus reinforce concerns about stability and reliance in arguing for thislimitation. Not because these concerns trump the goal of political satisfaction, but because that goal itselfrequires reliable estimates of political preferences that are sufficiently stable to affect behavior.

We would also expect to see (and do see) that courts are somewhat more prone to using a currentpreferences default rule – and less persnickety about how reliable the evidence of current enactablepreferences must be -- when enacting legislative preferences are particularly obscure or when the lapse intime since enactment is so great that the enacting polity likely has little preference at all about the future

171 See infra at VI (discussing cases where enacting legislative preferences are obscure or old).172 This is essentially the first case considered above.173 See CALABRESI, supra note , at 2, 6-7, 100-04.174 See ESKRIDGE, DYNAMIC, supra note, at 167-70; William N. Eskridge & John Ferejohn, Making the Deal Stick,

8 J. L. ECON. & ORG. 165, ___ (1992); William N. Eskridge, Jr. & John Ferejohn, The Article I, Section 7 Game, 80 GEO.L. J. 523, 549-51 (1992); Ferejohn & Weingast, supra note , at 267-68; Bill Eskridge, Overriding Supreme Court StatutoryInterpretation Decisions, 101 YALE L.J.. 331, 378-85 (1991); Schwartz, Spiller and Urbiztondo, A Positive Theory ofLegislative Intent, 57 LAW & CONTEMP. PROBS. 51, 56-57 & n.23 (1994). See also Linda Cohen & Matthew Spitzer,Solving the Chevron Puzzle, 57 LAW & CONTEMP. PROBS. 65, 69-76 (Spring 1994) (assuming there is a best s tatutoryinterpretation but that courts are free to deviate from it across an infinite continuum of views about policy and process);

53

event.171 If enacting preferences are highly unclear, even an uncertain estimate of current preferences ismore likely to increase political satisfaction.172 Likewise, even such an uncertain estimate of currentpreferences regarding present events they care about is more likely to maximize political satisfaction thana more accurate estimate of enacting preferences concerning events a century in the future they do not careabout at all. This is not because, as some have suggested, an old statute with a defined meaning can becondemned and overruled as obsolete by a court.173 It is rather based on the more limited ground that,when no defined statutory meaning exists, the enacting polity itself would prefer that enacting preferences,even when more ascertainable, should be discounted in the far distant future in favor of a default rule thattracks the preferences of the current polity, including the enacting polity’s own views about more ancientstatutes being interpreted during its own era.

In theory, one might adopt a full sliding scale to deal with this issue. Under this approach,interpretations should conform to current preferences if they can be ascertained with no less accuracy thanenacting preferences, but if current preferences can be ascertained less accurately, a court would weighthis greater inaccuracy against an implicit discount rate applied to future events. In practice, such a fullsliding scale seems unworkable, and instead courts understandably stick to a general rule of tracking currentenactable preferences when official action renders them reliably ascertainable, though with somewhatgreater willingness to find current preferences ascertainable the more obscure or ancient enactinggovernmental preferences are.

D. The Difference From Prior Rational Choice Models.The conclusions here diverge sharply from those in prior rational choice scholarship. This might

seem troubling to those (like me) who have been impressed the rigor and ingenuity of these models, whichhave made such great contributions to how we think about statutory interpretation. It thus may help toexplain the source of the divergence. As is often the case, differences in conclusion are traceable todifferences in assumptions, often assumptions that were asserted with little justification in the prior models.

To begin with, the rational choice models generally ignore any possible limits created by interpretivemethods, and instead assume judges always have discretion to reach whatever statutory interpretation theywant. They not only assume judges always have an interpretive choice, but generally assume judges makethat choice among an infinite policy continuum rather than being restricted by hermeneutic methodology toa limited set of plausible options.174 My assumptions are to the contrary, and I think more realistic given

Gely & Spiller, A Rational Choice Theory of Supreme Court Statutory Decisions with Applications to the State Farmand Grove City Cases , 6 J.L. ECON. & ORG. 263, 268-83 (1990) (basing most of analysis on assumption of infinitecontinuum but also considering possibility court can only make a yes-or-no choice). A slightly different take is offeredin Pablo T. Spiller, Rationality, Decision Rules, and Collegial Courts, 12 INT. REV. L. & ECON. 186, 188-90 (1992), whichassumes judges always have interpretive discretion to choose among any point in a policy space, and would generallyprefer to do so, but are sometimes driven by cycling problems to instead decide on a yes-or-no basis in order to achievestable agreement on the point most likely to come close to the judicial policy preferences. None of these articlesconsiders the possibility that hermeneutic methodology might reduce the interpretive choices to, say, three possibilities.Compare supra at IV (analyzing that possibility).

175 ESKRIDGE, DYNAMIC, supra note, at 167-70; Eskridge & Ferejohn, supra note , 8 J. L. ECON. & ORG. at ___; Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 548-51; Spiller, supra note , 12 INT. REV. L. & ECON. at 187-88; Gely& Spiller, supra note , at 267; Eskridge, Overriding, supra note , at 378-85; Schwartz, Spiller and Urbiztondo, supra note,at 56-57, 72; see also Cohen & Spitzer, supra note , at 71-76 (assuming those judicial views incorporate preferences aboutboth policy and legal process). Although they do not adopt this as their main assumption, Ferejohn and Weingastacknowledge this is the dominant assumption in rational choice literature, and adopt its as only alternative assumptionto possibility that judges try to further enacting preferences. Ferejohn & Weingast, supra note ,at 268-74. They thusdo not consider the possibility that judges might try to maximize current legislative preferences.

176 See, e.g., Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 548; McNollgast, Legislative Intent, supra note, at 5-6. The Landes and Posner model discussed above would fit in this category if it were extended to cases of fillingin default rules with enacting legislative preferences. See supra at __.

177 Ferejohn & Weingast, supra note ,at 268-74.178 Id.

54

the actual nature of legal reasoning. Judges have considerable interpretive choice, but it is not infinite andany accurate theory of statutory interpretation needs to incorporate that fact. Sometimes the hermeneuticanalysis dictates the interpretation and, even when it does not, usually it narrows the interpretive optionsto a handful.

More important, prior rational choice models virtually all assume that, in exercising whateverdiscretion they have, courts just try to maximize the satisfaction of judicial views to the extent they canwithout being overridden.175 I, in contrast, argue that courts actually try to maximize the satisfaction ofenactable political preferences, and in any event should so, and thus am interested in modeling what itwould look like if courts tried to fulfill that goal. Given the divergence on the two most fundamentalassumptions, the rational choice models not surprisingly have little connection to the analysis offered here.

A few rational choice models entertain the alternative assumption that judges try to act as an honestagent for the enacting legislature. Normally they leap to the conclusion that this assumption would meanjudges should interpret the statute to match enacting legislative preferences when they conflict with currentpreferences.176 In an ingenious twist, Ferejohn and Weingast argue that, while this is what a naive textualistwould so, a more politically sophisticated honest agent for the enacting legislature would realize that aninterpretation that matches enacting preferences might provoke an statutory override by the currentlegislature.177 They then conclude that a honest interpretive agency that seeks to maximize the satisfactionof enacting legislative preferences should first ascertain the range of interpretations that would not getoverruled by the current legislature, and then choose the option within that range that comes closest toenacting legislative preferences.178

179 Id. at 266-67 (collecting sources); Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 548-51. Because Ferejohnand Weingast focus only on future applications of the enacted statutes, and ignore present applications of all statutes,they conclude that durability is what would matter most to the enacting legislature. See supra at __.

180 Ferejohn and Weingast also never explain why any society would prefer to avoid legislative reconsiderationthat made clear where current enactable preferences lay. Rather than trying to avoid legislative reconsideration, in myanalysis courts can and should sometimes try to provoke such reconsideration when legislative preferences are unclear.See Elhauge, Preference-Eliciting, supra note .

181 ESKRIDGE, DYNAMIC, supra note, at 167-70; Cohen & Spitzer, supra note , at 69-70, 73-76; Eskridge &Ferejohn, supra note , 8 J. L. ECON. & ORG. at ___; Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 528-51; Ferejohn& Weingast, supra note , at 267-76; Eskridge, Overriding, supra note , at 378-85; Gely & Spiller, supra note , at 267-83.

182 ESKRIDGE, DYNAMIC, supra note, at 164-70; Cohen & Spitzer, supra note , at 68; Eskridge & Ferejohn, supranote , 8 J. L. ECON. & ORG. at ___; Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 549. Often this assumption is notexplicit but just an implicit premise of how the model is drawn and applied. See, e.g., Pablo T. Spiller, Rationality,Decision Rules, and Collegial Courts, 12 INT. REV. L. & ECON. 186, 187-90 (1992); Ferejohn & Weingast, supra note ,at 267-76; Gely & Spiller, supra note , at 270-83. The one exception is Schwartz, Spiller and Urbiztondo, A Positive Theoryof Legislative Intent, 57 LAW & CONTEMP. PROBS. 51 (1994), but that paper makes various other odd assumptions thatlimit its utility. It assumes the only purpose of textual specificity and legislative history is to signal a likelihood oflegislative override. Id. at 55-56, 60-61, 64-69. Further, its assumption that courts just try to maximize judicial preferencesleads them to the conclusion that legislatures will want to send false signals to the judiciary. Id. at 64-70. Theseassumptions seem erroneous and are certainly contrary to mine. I address other oddities with their model in Elhauge,Preference-Eliciting, supra note , at II.

55

Ferejohn and Weingast are led to a conclusion that differs from mine because they share anotherassumption with other rational choice models, an assumption so taken for granted that it is just left implicit.That assumption is that the only legislative preferences to further are those reflected in the statutes enactedby that legislature, and that thus enacting legislative preferences are maximized by achieving future statutoryresults that come as close as possible to enacting preferences.179 If one instead asks which general defaultrule the enacting legislature would prefer for both old and new statutes, the inquiry produces instead theconclusion above: that the enacting legislature itself would want a default rule that tracks current legislativepreferences (where reliably ascertainable) because it cares not just about the future but about the presentduring which it is in office. Thus, in my analysis, unlike the Ferejohn and Weingast model and other priormodels, a honest interpretive agent for the enacting legislature would choose a current preferences defaultrule.180

Rational choice models of statutory interpretation also tend to focus on a particular legislativestructure – normally the division between Presidential, House and Senate authority in the U.S. System, andsometimes even the particular legislative committee structure.181 My aim here is instead to develop aninterpretive theory for general application across different legislative structures, including Parliamentary onesthat do not involve a similar separation of legislative powers.

Finally, prior rational choice models assume that executive and legislative preferences are perfectlyknown.182 This assumption seems entirely unrealistic. In contrast, I assume that many interpretative canonsreflect judicial uncertainty about those preferences, and can best be explained as attempts to rely on indirectmeans to estimate or elicit them. This difference in assumptions will be particularly central in explaining how

183 See infra Part VII.184 See Elhauge, Preference-Eliciting, supra note .185 See, e.g., Gely & Spiller, supra note , at 266 ("Congressional inaction will follow Supreme Court decisions,

as these have already taken the composition of Congress into account.")186 See Elhauge, Preference Eliciting, supra note , at II.187 Id. at II, V.C.188 Id. at II.B.3.189 See sources cited supra note .

56

the Chevron doctrine helps courts estimate current legislative preferences,183 and how other canons helpcourts elicit legislative preferences when they are uncertain.184 Indeed, if this assumption of perfectly knownlegislative preferences were true, most statutory canons and default rules would be unnecessary.

I will argue below that actual doctrine is more consistent with the proposition that courts follow nottheir own preferences nor just enacting legislative preferences but instead follow current preferences defaultrules when that yields reliable estimates of legislative preferences. But there is also some more generalempirical evidence that bears on the question. Prior rational choice models essentially predict that courtswill pay attention to current preferences only to the extent necessary to avoid legislative override. Giventheir assumption that courts have perfect knowledge of legislative preferences, these models would predictjudicial interpretations are never overridden.185 In fact, statutory overrides occur with surprising frequency,which my theory explains on the grounds that, when legislative preferences are uncertain, the courts oftentry to elicit legislative reactions.186 If one instead adopts the more realistic assumption that courts haveimperfect knowledge and thus sometimes mistakenly overplay their hand, these rational choice models stillface empirical difficulty. For then one would think courts would at least avoid using statutory canons thatare known to be especially prone to eliciting statutory overrides. But in fact courts use those canons andeven sometimes expressly call for statutory overrides.187 Further, since these models predict courts willtry to interpret statutes in ways favored by a sufficient legislative faction to block override, one would thinkthat, when the judges miscalculate, the resulting statutory overrides would be partisan and controversial.In fact, most overrides are neither.188

E. Limits on a Current Preferences Default RuleUnder my analysis, it seems likely that the political preferences of the enacting government – and

a fortiori those of the current government – would be maximized by a general default rule that, in casesof statutory ambiguity, tracked current enactable preferences when those preferences can reliably beascertained. But some important limitations on this conclusion must be stressed.

First, because this default rule does not rely on the ground that current governmental preferencesshould take precedence over those of the enacting government, it does not justify the position (taken byvarious prominent scholars), that current preferences should modify statutory meaning itself.189 Theproposition here rests on the much more supportable ground that -- where statutory meaning is unclear --the enacting government itself would agree on a general default rule that maximizes the preference

190 See supra Section C.191 See Elhauge, Preference-Eliciting, supra note , at XI.B.192 See supra at II-III; Elhauge, Preference-Eliciting, supra note , at XI.C.

57

satisfaction of each government during its tenure in office. Moreover, like any default rule, this one onlyoperates within the range of plausible statutory meanings left by the original enactment.

Why not allow courts to change even unambiguous statutory meaning to conform to current politicalviews? Strictly speaking, my thesis permits me to remain agnostic on this question. It suffices to say that,whether or not one views statutory meaning as altered by changed political preferences, the statutorydefault rule used to resolve statutory ambiguities should be changeable. But if my argument extended tochanging statutory meaning itself, then the case of statutory ambiguities would simply be a special case ofa more radical thesis, so it bears explanation why I do not take my analysis that far, and why my reasonsfor stopping short do not extend as well to resolving statutory ambiguities.

To begin with, my rebuttals to objections to a current preferences default rule depended in part ofthe existence of a statutory ambiguity to explain why the enacting polity did not have a special interest, orwould not prefer greater durability, in the statutes it enacted.190 Likewise, my response to the argumentthat reliance interests should bar a current preferences default rule turns critically on the premise that therewas no clear statutory meaning on which reasonable reliance could be made.191

More important, an unambiguous statutory meaning embodies the clearest indication of what wasactually able to produce legislation. The current government has not in fact enacted a statute. It thus hasnot gone through the political process necessary to make their pronouncements authoritative -- such as,in the United States, either winning a majority in two legislative chambers and Presidential approval, orovercoming a Presidential veto with a two-thirds vote in two chambers. Only the enacting government has,and thus if we ask what meaning to attach to the statute, we must ask what meaning to ascribe to theirauthoritative action.

The current government might want to change that meaning, but that proposition should be testedby having it actually do so through the same means that the enacting government used. Every politicalsystem has elaborate procedures for enacting statutes to make sure that a considered choice was made thatreflected prevailing political preferences. Sometimes these seem inconvenient. But if these procedureshave become too cumbersome, then they should be modified to allow more nimble entities to take action,or the legislature should delegate ongoing lawmaking power. Action by the enacting legislature thatcompleted the constitutionally required process cannot be reversed simply because a judge believes thecurrent government would probably be able to complete that same process with a different result. Thatwould circumvent the constitutionally required procedure for enacting statutes, and allow the currentgovernment to use a less exacting procedure to undo what the earlier government did.

In contrast, a current preferences default rule does not circumvent legislative procedure becausethe ambiguity has been left by a duly enacted statute that was nonetheless intended to cover the matter.When resolving statutory ambiguities, traditional and constitutionally authorized means of statutoryinterpretation often call for courts to refer to extrinsic evidence of legislative preferences, even though thatevidence did not itself go through the enactment process.192 Referring to such evidence regarding the

193 See infra Parts VI-VII.194 Lon L. Fuller, The Morality of Law 209-10 (rev. ed. 1969).195 Special issues might be raised when a prior judicial interpretation exists, and are discussed Elhauge,

Preference-Eliciting, supra note , at XI.B.

58

current legislature no more circumvents constitutional enactment procedure than referring it for the enactinglegislature. Nor does it permit the current legislature to undo what the enacting legislature did through a lessexacting procedure. To the contrary, as we shall see, courts tend to be more exacting in the reliability ofevidence of current preferences they will consider than they are for enacting preferences.193

Even if we thought courts had authority to deviate from statutory meaning to maximize politicalsatisfaction, the enactable preferences of the current legislature (which enacted no statutory meaning) arenecessarily less susceptible of reliable estimation than those of the enacting legislature that actually enacteda statutory meaning to govern the issue. Extending the current preferences approach to statutory meaningwould thus introduce far more error because judges will often be wrong about what the current governmentwould enact. A considered definite choice by the enacting should not be overturned simply because ajudge believes the current government probably would do so.

In contrast, using a current preferences default rule to resolve statutory ambiguity, no new form oferror is introduced since the ambiguity already forces judges to guess about what some government wouldhave wanted. Judges’ ability to accurately estimate how a government would have wanted to resolve theuncertainty may indeed often be stronger for the current government than for the enacting government.While the enacting government did enact a statute on the subject at hand, by definition it left no clearmeaning on the issue at hand. Moreover, any expression of the current government’s views are more easilyinterpretable because they are current with the interpreter and thus not thrown into doubt by changes infactual circumstances or linguistic conventions. In any event, courts tend to restrict themselves to relativelymore reliable evidence of current preferences than they use to estimate enacting preferences.

When statutory meaning is discernable, following it is also required by standard rule of law norms.As Lon Fuller observed, the rule of law requires that "in acting upon the citizen... a government will faithfullyapply rules previously declared as those to be followed by the citizen and as being determinative of hisrights and duties."194 A defined statutory meaning constitutes such a declaration. While ambiguities instatutory interpretation leave some matters to judicial judgment, that is no excuse to expand that power tocover all topics, including those resolved by an accepted statutory meaning. Cabining judicial discretionat the point of application, when the identity of the parties benefitted or hurt is known, is to the contrary animportant part of the rule of law.

In contrast, where the meaning of a duly enacted statute is unclear, then judges have no choice butto make some decision about what default rule to choose to resolve that unclarity. And such interpretationdoes not violate rule of law norms because there was no unambiguous statutory meaning which citizenscould know to follow.195 To the contrary, the issue is by definition one that otherwise would be left tojudicial judgment, and the default rule here is just a way of cabining that judgment.

Finally, where the enacting government has adopted a defined statutory meaning, it has made clearthat it does want its political preferences regarding that meaning followed. In contrast, in deciding on the

196 See Bundy & Elhauge, Knowledge About Legal Sanctions, 92 M ICH. L. REV. 201 (1993).197 True, that future legislature could itself amend the rule of interpretation, but the issue may never reach the

legislative agenda, or that future legislature may not be able to agree on an alternative rule with enough consensus toovercome procedural obstacles and effective supermajority requirements.

59

best statutory default rule, following the preferences of the current government (where they can reliably bedetermined) does not thwart the choice of the enacting government. Rather, it conforms to the default rulethat the enacting government would itself choose.

Second, as with all default rules, a current preferences default rule is subject to opt-out. That iswhat makes it a default rule rather than a mandatory rule. This provides another reason why courts shouldnot modify explicit statutory meanings to conform to changes in legislative preferences, for adopting adefined statutory meaning is how governments opt out of statutory default rules. Modifying explicit statutoryterms would take away the governmental power to opt out of a current preferences default rule.

True, sticking with an unambiguous statutory meaning will sometimes produce what even theenacting government would regard as an undesirable result. But, by adopting explicit terms that opt outof the judicial process of filling and updating default rules, the enacting government clearly indicated thatit viewed the error costs of following an inflexible explicit term to be lower than the error costs of judicialadaptation. The enacting government effectively faced a choice between a rule and a standard -- requiringa decision about whether the inherent over- and underinclusion of a fixed rule is worse than the over- andunderinclusion caused by erroneous application of a less precise standard.196 This is a choice on whichreasonable persons can differ, and adopt different conclusions for different areas, and it is a choice theenacting government is entitled to make. In contrast, a default rule that (where statutory meaning isambiguous) changes to accommodate changing political preferences does not thwart a government’s choiceto opt out of judicially set default rules.

But my analysis also suggests limitations on the opt out. A government can opt out by adoptingdefined statutory meanings. But plainly a government cannot enact a statute providing that its preferencesgovern both the interpretation of ambiguities in older statutes and the interpretation of its enacted statutesin the future. That is, a government cannot simultaneously provide that a current preferences default rulegoverns past statutes and an enacting preferences default rule governs its statutes when interpreted in thefuture. To do so would be to allow the government at one point in time to aggrandize its power relativeto the power of past and future governments.

A more problematic case is where a government enacts a general interpretive statute providing that,from now on, statutory ambiguities should be interpreted to further the preferences of the enactinggovernment. If applied both to older statutes interpreted during its time in office and to the statutes itenacted that are interpreted during the future, such a trade off might seem unproblematic. But if such achange in interpretive rules were enacted toward the end of the government’s reign, that government wouldin effect have had the benefit of a current preferences default rule during its time in office, and then beimposing an enacting preferences default rule for the interpretation of its statutes in the future. Moreover,such a statute would also apply to future governments that might have different views about the best defaultrule for statutes interpreted or enacted during its reign.197 Courts would then have to decide between

198 See generally Vermeule, Interpretive Choice, 75 N.Y.U. L. REV. 74, 97 n.91 (2000) (“the scope of legislativeauthority to prescribe interpretive rules is not clear in a system of separated powers”); Elhauge, Preference-Eliciting,supra note , at XI.C (noting that power of interpretation is vested by Article III on courts).

199 See Elhauge, Preference-Eliciting, supra note .

60

applying the current preferences default rule that seems likely to maximize that future government’s politicalpreferences, or using the enacting preferences default rule preferred by a past government that did not enactthe future (or older) statutes being interpreted and does not represent current legislative preferences duringthe time of interpretation. Such an instruction about how courts should interpret older and futuregovernment’s statutes in the future not only thwarts the power of future governments (and politicalsatisfaction generally) but also arguably violates the constitutional clauses vesting interpretive power in thejudiciary.198 One might accordingly predict that such opt outs will be treated with hostility. Indeed, evenif doable, it may well be that such an opt out cannot be effectuated by one legislature but rather requiresa constitutional amendment to bind future legislatures and polities and modify the constitutional power ofjudges.

Third, any current preference default rule only matters in the marginal area where explicit actionby the current government is unlikely. If the current government felt strongly enough about the conflictbetween its preferences and past legislative preferences, it could always enact explicit legislation to thateffect. It is only when this conflict is not strong enough to overcome legislative inertia and motivate currentlegislative action that the default rule actually affects the outcome. In short, it takes a lack of explicitlegislation by both the enacting and the current government to create the situation in which a currentpreference default rule governs.

This point reveals the error in the following sort of objection: that the enacting government wouldnot prefer a current preferences default rule because, if it really did not like the interpretation of any pastlegislation during its time in office, it could always override it. Although it is true a legislature can alwaysoverrule any interpretation it really does not like, it is also true that the enacting government would realizethat any default rule (such as an enacting preference default rule) can only govern in the future if futuregovernments are not sufficiently motivated to override it. Thus, the real choice at each point in time is whatdefault rule will govern in this marginal area where explicit legislative action is unlikely by either legislature.In that marginal area, the enacting government would prefer to have its political preferences govern duringits time in office, rather than have old preferences govern during its time and its preferences govern duringthe future. In any event, the analysis so far puts aside the possibility of legislative override, which I considerat length in a companion piece.199

Fourth, the political preferences relied upon to provide content to the current preference defaultrule must be the preferences of a set of political actors who could actually enact legislation. The courtsshould not use the leeway provided by a current preference default rule as an excuse to favor the politicalpreferences of one political party over the other, when one controls the executive branch and the other thelegislature, or each control different legislative chambers. In such a case, the political preferences of eitherparty do not suffice to enact legislation and neither should be taken as an appropriate measure of currentenactable preferences.

200 See supra at 201 Infra at Elhauge, Preference-Eliciting, supra note , at XI.B (rejecting claims that reliance interests should

trump political preference satisfaction but explaining the extent to which political satisfaction would be advanced byinterpretations that are sufficiently stable to induce some reliance).

61

This is an important difference from prior theories of statutory interpretation that have arguedcontemporary values ought to be allowed to update old statutes.200 Such theories view judges more aspartners than as agents, picking among current majority preferences to determine which reflect soundcontemporary values that should govern statutory interpretation even if those preferences are not enactable.The theory here, instead, views judges interpreting statutes as agents for the set of political preferencespossessed by the polity that could actually secure enactment.

The enacting legislative polity would not want a default rule that tracked nonenactable politicalpreferences because that would not maximize the preference satisfaction of those political interests that canenact legislation. Governments would not want to empower courts to take sides where political gridlockexists, rather than just serving as honest agents for the political forces that can command enough politicalagreement to enact statutes. That would just expand the influence of judicial preferences over politicalones. Instead, governments would want courts to rely on current enactable preferences. In cases wherethe two political parties are at loggerheads on the relevant issue — and either one can veto change by theother— the court must rely instead on the most recent indication it had of political preferences that sufficedto enact legislation.

Even if the legislative polity did want courts to track nonenactable preferences, that would beconstitutionally problematic. For example, in the United States, the bicameralism and presentment clauseswere meant to bar laws that could not secure sufficient approval under such a system. These clauses givesa minority of the population in less populous states a veto over the majority of the population, gives themajority of the population a veto over the majority of the states, and gives an nationally elected official aveto over legislators elected by district or state. Each political actor can thus insist that laws advance theinterest of its constituency. Although we are within the realm of statutory interpretation where some judicialjudgment about the proper default rules is required, for judges to exercise that judgment to further politicalpreferences that could not be enacted would undermine this constitutional structure.

Fifth, especially where statutory precedent exists, a change in enactable political preferences canjustify a change in interpretation only if the changed preferences are memorialized in some relatively well-defined official political action. It cannot be that every dip up or down in the polls, or changed reading ofthe political tea leaves, alters statutory interpretation. Such an unstable legal regime would fail to inducethe behavioral reliance that is necessary to make interpretations effective enough to advance any politicalpreferences.201 It might even violate rule of law norms by creating massive uncertainty about legalconsequences and making notice and planning impossible. Moreover, officially indicated changes in currentpreferences will be necessary to make sure those enactable preferences have reliably been ascertained.Enactors will want fairly high standards of reliability for they will recognize that judges are agents, and thatgiving judges open-ended interpretative power thus creates not only error costs (good faith errors inguessing about changing legislative preferences) but also agency costs (furthering judge’s personal

202 See infra VI.C.203 This would also apply to legislative decisions to let stand agency interpretations that either were decided

before Chevron or fall outside its scope. See infra at VII (discussing Chevron deference).204See United States v. Rutherford, 442 U.S. 544, 554 n.10 (1979); Red Lion Broadcasting Co. v. FCC, 395 U.S.

367, 381-382 n.11 (1969). Indeed, such legislative inaction might signal a failed preference-eliciting default rule, and thusaffirmatively justify rejecting the prevailing interpretation. See Elhauge, Preference-Eliciting, supra note .

62

preferences in the guise of following current legislative views). Because current preferences can often beestimated with less accuracy than enacting preferences, a current preferences default rule best reduces errorif confined to cases where those current preferences are reliably ascertainable. Even if error costs are thesame when making estimates about past and current governments, agency costs are increased by allowingjudges to alter interpretations with changing legislative preferences. For under such an approach, statutoryprecedent would be far less binding, and judges would be tempted to change interpretations as judicialpersonnel changed. Instead, statutory precedent should remain binding until the change in enactablepolitical preferences has been memorialized in some official action. As we will see in the next two sections,that limitation largely reflects the actual judicial practice.

VI. TRACKING CURRENT LEGISLATIVE PREFERENCES IN PRACTICE

Many judicial doctrines have long seemed unjustifiable if designed to inquire into the enactinglegislature’s meaning, intent, or preferences. But the next sections show that many of these doctrines canreadily be justified as current preference default rules. Moreover, the theoretical limits on currentpreference default rules may help explain what has often been criticized as the inconsistent application ofthese doctrines.

A. Subsequent Legislative Action Retaining or Relying on Otherwise NonbindingInterpretations.

When a higher court has interpreted a statute, lower courts are bound by that interpretation untilthe higher court changes its mind. When a court has itself interpreted the statutory provision, that precedentmight be considered binding under stare decisis.202 But often courts face the question whether to fill instatutory ambiguities with interpretations by lower or coordinate courts that subsequent legislaturesaffirmatively decided to let stand.203 Should courts follow those interpretations even if they do not seemthe best reading of the enacting legislature’s preferences?

Under the prevailing law, the mere fact of subsequent legislative inaction in the face of a prevailinginterpretation is not sufficient to make the outstanding interpretation persuasive to a coordinate or highercourt.204 But once the interpretation “has been ‘fully brought to the attention of the public and theCongress,’ and the latter has not sought to alter that interpretation although it has amended the statute in

205 Rutherford, 442 U.S. at 554 n.10; North Haven Bd. of Educ. v. Bell, 456 U.S. 512, 530-35 (1984); see alsoCommodity Futures Trading Comm'n v. Schor, 478 U.S. 833, 846 (1986) ("It is well established that when Congress revisitsa statute giving rise to a longstanding administrative interpretation without pertinent change, the 'congressional failureto revise or repeal the agency's interpretation is persuasive evidence that the interpretation is the one intended byCongress."') (quoting NLRB v. Bell Aerospace Co., 416 U.S. at 274-75); see also Bob Jones University v. United States,461 U.S. 574, 599-602 (1983); Herman & MacLean v. Huddleston, 459 U.S. 375, 384-86 (1983); Merrill Lynch, Pierce, Fenner& Smith, Inc. v. Curran, 456 U.S. 353, 379-382 (1982); Haig v. Agee, 453 U.S. 280, 300-301 (1981); Cannon v. Universityof Chicago, 441 U.S. 677, 686 n.7, 702-03 (1979).

206 See, e.g., United States v. Board of Comm'rs, 435 U.S. 110, 134-35 (1978) (“When a Congress that re-enact sa statute voices its approval of an administrative or other interpretation thereof, Congress is treated as having adoptedthat interpretation, and this Court is bound thereby. . . . While . . . it is impermissible to draw inferences of approval fromthe unexplained inaction of Congress, . . . that principle has no applicability to this case. ”) (collecting cases); Lorillardv. Pons, 434 U.S. 575, 580 (1978); Shearson/American Express v. McMahon, 107 S. Ct. 2333, 2343 (1987); Helvering v.Hallock, 309 U.S. 106, 130-32 (1940).

207 See, e.g., FDA v. Brown & Williamson, 120 S. Ct. 1291, 1312-13 (2000) (“We do not rely on Congress' failureto act . . . this is not a case of simple inaction by Congress . . . To the contrary, Congress has enacted several statutesaddressing the particular subject . . . “); Runyon v. McCrary, 427 U.S. 160, 173-75 (1976); Apex Hosiery v. Leader, 310U.S. 469, 488-89 (1940).

208 See, e.g, HENRY M. HART , JR. & ALBERT M. SACKS, THE LEGAL PROCESS 1284-86 (tentative ed. 1958).209 Where it is close it time, it might be relevant to ascertaining enacting preferences, see supra at III.B.210 See Eskridge, Overriding, supra note , at 397, 400-03 (agreeing that court invokes “legislative inaction and

subsequent legislative history as evidence of statutory meaning when it believes that they provide reliable evidence ofcurrent congressional preferences”).

63

other respects,” then that interpretation is presumptively correct.205 The same follows when the legislatureallows an interpretation to stand when it re-enacts a statute206 or enacts other statutes on the samesubject.207 Legislative action that consciously lets an interpretation stand suffices, whereas mere legislativeinaction would not.

This reliance on subsequent legislative action is hard to square with a hermeneutic interpretation ofthe original statute. The actions of a subsequent legislature are irrelevant to the meaning or intent of theenacting legislature.208 Similarly, subsequent legislative action that comes significantly later in time isirrelevant to estimating the political preferences of the enacting legislature,209 and thus cannot be explainedas an enacting preferences default rule. Nor has the subsequent legislation enacted any amendmentregarding the issue at hand that would itself have the force of law or create a new statutory meaning.

But this caselaw does make sense as a current preferences default rule, for the subsequentlegislative action provides a good indication of more current enactable preferences.210 Consistent with thisapproach, courts are not willing to rely on the sort of subsequent legislative events that fail to provide anyreliable means of ascertaining enactable preferences. Mere legislative inaction cannot offer such a reliablemeans, for inaction can result from lack of time or attention to the issue, or from an inability to gain theconcurrence of some key committee, the other legislative house, or executive branch. Neither inattentionnor deadlock can indicate current enactable preferences with sufficient reliability to justify deviating from

211 In contrast, proof of such legislative inaction can rebut a claim that current enactable preferences havechanged in a way that justifies deviating from a prevailing interpretation that does conform to enacting legislativepreferences. See infra Section VI.C.

212Aaron v. SEC, 446 U.S. 680, 694 n. 11 (1980) (rejecting reliance on amendments “addressed principally tomatters other than that at issue here”); Zuber v. Allen, 396 U.S. 168, 185 n.21, 192-94 (1969); Monell v. New York CityDept. of Social Servs., 436 U.S. 658, 696-99 (1978); Girouard v. United States, 328 U.S. 61, 70 (1946); Leary v. United States,395 U.S. 6, 24-25 (1969).

213 See, e.g., Franklin v. Gwinnett County Public Schools, 503 U.S. 60, 78 (1992) (Scalia, J., concurring in thejudgment); FDA v. Brown & Williamson, 120 S. Ct. 1291, 1306-13 (2000).

214Compare ESKRIDGE, DYNAMIC, supra note , at 139.215 Pension Benefit Guaranty Corporation v. LTV Corp., 496 U.S. 633, 650 (1990); United States v. Price, 361 U.S.

304, 313 (1960); Mackey v. Lanier Collections Agency & Serv., 486 U.S. 825, 840 (1988); Public Employees Retirement

64

the likely preferences of the enacting legislature.211

But where the subsequent legislature took the time to amend, enact, or re-enact a statutoryprovision, without disturbing an outstanding interpretation that had been brought to its attention, then thatoutstanding interpretation does become a fairly reliable indicator of the balance of political forces affectinghow the more current legislature would want the default rule filled. And the fact that the more currentlegislature took affirmative legislative action means that the other house and executive concurred in thedecision to let the interpretation stand, so that the preferences implied by that action are enactable.

The persuasive objection to relying on subsequent legislative inaction -- that it might reflect a lackof time, attention, or agreement rather current political preferences -- also applies to affirmative legislativeaction if the acting legislature was unaware of any prevailing interpretation, or believed (accurately orinaccurately) that no single interpretation prevailed. The objection would also still apply if the subsequentlegislature was aware of a prevailing interpretation, but its legislative action focused on unrelated provisionsof the statute. Consistent with this, courts in fact do not rely on subsequent legislative action to justifydeviating from enacting legislative preferences in cases meeting these descriptions.212

Sometimes subsequent legislatures affirmatively rely on a particular statutory interpretation inenacting another statute. In such cases, even justices otherwise loathe to draw inferences from subsequentlegislative action agree the relied upon assumption should be adopted.213 The reason is not, I will arguein Part V, because mistaken reliance is generally a sensible ground for a default rule.214 It is, rather,because the reliance by the subsequent Congress provides a fairly reliable indication that more currentpolitical preferences favor the interpretation in question.

B. Other Evidence of Changed Legislative PreferencesWhat should a court do when subsequent legislative history or statutes do not retain or rely on a

prevailing interpretation, but more generally indicates that the current government would resolve a matterdifferently from the enacting government? The U.S. Supreme Court has sometimes seemed inconsistentin its treatment of such subsequent legislative history or statutes. On the one hand, the Court has stressedthat: "the views of a subsequent Congress form a hazardous basis for inferring the intent of an earlierone."215 On the other hand, the Supreme Court has held that: "while the views of subsequent Congresses

System v. Betts, 492 U.S. 158 (1989); Consumer Prod. Safety Comm'n v. GTE Sylvania, Inc., 447 U.S. 102, 117-18 (1980).See also Haynes v. United States, 390 U.S. 85, 87-88, n. 4, 88 S.Ct. 722, 19 L.Ed.2d 923 (1968) (“The views of a subsequentCongress of course provide no controlling basis from which to infer the purposes of an earlier Congress.”); United Statesv. Southwestern Cable Co., 392 U.S. 157, 170 (1968) (“the views of one Congress as to the construction of a statuteadopted many years before by another Congress have ‘very little, if any, significance.’”); Sullivan v. Finkelstein, 496 U.S.617, 632, 110 S.Ct. 2658, 110 L.Ed.2d 563 (1990) (SCALIA, J., concurring) ("Arguments based on subsequent legislativehistory ... should not be taken seriously, not even in a footnote.").

216 Seatrain Shipbuilding Corp. v. Shell Oil Co., 444 U.S. 572, 596 (1980); Andrus v. Shell Oil Co., 446 U.S. 657,666 n.8 (1980).

217 Morton v. Mancari, 417 U.S. 535, 549 (1974) (collecting cases). 218 FDA v. Brown & Williamson, 120 S. Ct. 1291,1306 (2000) (quoting United States v. Estate of Romani, 523 U.S.

517, 530-31 (1998)). See also id. at 1301 (“the meaning of one statute may be affected by other Acts, particularly whereCongress has spoken subsequently and more specifically to the topic at hand.”);United States v. Fausto, 484 U.S. 439,453 (1988) (“the implications of a statute may be altered by the implications of a later statute.”); Bowen v. Yuckert, 482U.S. 137, 148-52 (1987) (interpreting Social Security Act in light of Social Security Disability Benefits Reform Act of 1984).

219See, e.g, HART . & SACKS, supra note, at 1284-86 (treating legislative history as irrelevant unless “officiallybefore the legislature at the time of its enactment”).

220120 S.Ct. at 1306.

65

cannot override the unmistakable intent of the enacting one, such views are entitled to significant weight andparticularly so when the precise intent of the enacting Congress is obscure."216 In some cases, the Courthas recited — as a “cardinal rule” no less — that “repeals by implication are not favored,” thus indicatingthat Congress must be explicit when it repeals prior legislation217 In other cases, the Court states that: “aspecific policy embodied in a later federal statute should control our construction of the [earlier] statute,even though it ha[s] not been expressly amended."218

In short, the Court has sometimes been willing to be guided by subsequent legislative history,especially when it takes the form of statutes that (even though not repealers) indicate current enactablepreferences that bear on that topic. This would be puzzling as a matter of hermeneutics. If the meaningof the enacting legislature were the touchstone, the views of subsequent legislators expressed even in officiallegislative history should be irrelevant.219 Nor would the subsequent enactment of (non-repealing) statutesbear on the original legislature’s meaning.

But none of this matters if the interpreter is actually just applying a current preferences default rulein a case where the enacting government’s meaning is unclear. The views of the subsequent legislature mayindeed be a “hazardous basis” for inferring the meaning or intent of an earlier legislature, but they are notbeing used for that purpose. They are, rather, being used to fill the default when the enacting government’smeaning and even preferences are “obscure.”

Likewise, implied repeals are suspect when there is an enacted statutory meaning that has gonethrough the political process, but not when the prevailing interpretation is just a default rule that estimatesenactable preferences to plug a gap in statutory meaning. The courts accordingly hold that the doctrineallowing statutory interpretation to be altered by subsequent legislation applies only when at the time ofenactment, the statute has “a range of plausible meanings.”220 If that is the case, “subsequent acts can shapeor focus those meanings,” that is they can provide a basis for selecting an interpretive option among the

221 Id. (emphasis added).222 See supra Section IV.A (explaining why both relative clarity and age of enacting preferences would matter

under a current preferences default rule).223 Bob Jones University v. United States, 461 U.S. 574 (1983). 224 Id. at 593 n.20.

66

range of plausible statutory meanings.221 In short, we have a doctrine against implied repeals of actualstatutory meaning but not against implied repeals of default rule interpretations that fill gaps in meaning.

The remaining complication is that subsequent legislative history is often a hazardous basis forinferring current enactable preferences as well. It is particularly difficult to ascertain how much weight togive to legislative statements that were not memorialized in, or did not lead to, the enactment of actuallegislation. Through enactments a legislature speaks in one voice, and manifests the political preferencesthat sufficed to secure the actual enactment of something. In other statements, legislatures speaks throughmultiple voices, none of which may reliably indicate enactable preferences. Some skepticism is warranted.Which is why the Court quite properly restricts use of subsequent legislative history to cases where eitherenacting preferences are particularly obscure or current enactable preferences are especially clear. Theparadigm case is the enactment of a subsequent (non-repealing) statute that nonetheless reliably indicatesthe prevailing political preferences most likely to be able to secure enactment in the current legislature. Anew statute is no more reliable an indicator of the earlier government’s meaning, intent, or preferences thanmere statements by current legislators would be. But a new statute is a more reliable indicator of currentenactable preferences and thus has greater influence on judicial interpretation of earlier statutes.

This analysis means that courts cannot be deemed internally inconsistent merely because theysometimes rely on subsequent legislative history and statutes and sometimes do not. Subsequent legislativehistory and statutes should indeed be rejected when the enacting legislative intent or meaning is clear.Moreover, even when the original meaning is unclear, subsequent legislative history should be rejectedwhen it provides an unreliable indicator of current enactable legislative preferences. But when it doesprovide a reliable indicator of current preferences, particularly as reflected in subsequently enacted statutes,then that subsequent legislative history offers sufficient grounds for filling the default. And the more obscureor ancient the enacting preferences, the less demanding the courts will be about how reliable the indicationof current preferences must be.222

For example, the Supreme Court in Bob Jones concluded that religious nonprofit universities withracially discriminatory admissions standards were not “charitable” organizations entitled to tax exemptionunder federal statute.223 The Court conceded that the 1894 legislature that enacted the statute probablywould not have regarded such admissions standards as problematic given the prevalence of racialsegregation at the time.224 If one asks the question what the enacting legislature would have wanted, orwhat its political preferences were, the most likely answer would be to grant the exemption.

But by 1983, it was clear that, while the current Congress had not decided the precise question,granting such a charitable exemption fundamentally contradicted modern congressional race relations policy.The Court could have reached this conclusion by relying simply on congressional acquiescence in an IRS

225 461 U.S. at 599-602 (noting that Congress had rejected 13 bills to overturn the IRS interpretation despitemaking other enactments, including one that denied any tax exemption to racially discriminatory social clubs). See supraSection IV.C.1. This IRS interpretation might also seem to merit deference under tax caselaw that preceded the Chevroncase. See BobJones, 461 U.S. at 596-97. But this theory faced the considerable difficult that the then-current TreasuryDepartment had announced its intent to revoke this interpretation. 461 U.S. at 585 n.9. See generally Section D(discussing how Bob Jones fits Chevron deference).

226Id. at 593-94.227 Id. at 598.228 461 U.S. at 585 n.9. 229 See Legislation to Deny Tax Exemption to Racially Discriminatory Private Schools, Hearing Before the Senate

Committee on Finance, 97 Cong. 2d Sess., at 225-26, 229 (Feb. 1, 1982) [hereinafter “Tax Exemption Hearing.”]. For anexplanation of why the court should not defer to such an agency interpretation under Chevron, see infra Section D.

230 See Tax Exemption Hearing at 225, 231-32.231 See Tax Exemption Hearing at 244, 247.

67

interpretation that Congress had repeatedly refused to alter even though it otherwise amended the statute.225

But the Court first made clear that the same conclusion would follow (even without any legislativeacquiescence in a prevailing interpretation) because: “Over the past quarter of a century, everypronouncement of this Court and myriad Acts of Congress and Executive Orders attest a firm nationalpolicy to prohibit racial segregation and discrimination in public education.”226 Further, when the Courtlater approved the IRS interpretation, the Court did so not because the IRS correctly divined what the1894 Congress would have wanted, but because the IRS correctly estimated current legislativepreferences. “In 1970, when the IRS first issued the ruling challenged here, the position of all threebranches of the Federal Government was unmistakably clear. . . . Indeed, it would be anomalous for theExecutive, Legislative and Judicial Branches to reach conclusions that add up to a firm public policy onracial discrimination, and at the same time have the IRS blissfully ignore what all three branches of theFederal Government had declared.”227 The clear import was that both the courts and agencies should,in case of ambiguity, follow the clearly indicated preferences of the current legislature, here reflected inrepeated enactment of actual laws.

A skeptic might wonder whether these enactments really reflected current enactable preferencesin 1983 given that the then-current Reagan administration had, through its Treasury Department, announcedits intent to revoke the prevailing IRS interpretation against granting tax exemptions to raciallydiscriminatory schools.228 However, the Reagan Treasury Department made plain this recission reflectedits legal interpretation, not its political views.229 Indeed, the Reagan administration simultaneouslyintroduced a bill before Congress to statutorily deny any tax exemption to racially discriminatory schools.230

Democrats alleged that this bill was just a response to the political outcry that attended the administrationdecision to rescind the IRS interpretation,231 but that is besides the point. What matters are not the privatepolitical preferences of political actors, but what their actions show about the state of current enactablepreferences. If political actors who suffice to enact legislation would be unable to resist political pressurefor a certain result, then that suffices to show that result reflects an enactable political preference, regardlessof the private preferences those political actors might have.

232 Tax Exemption Hearing at 234-35, 253. Chairman Dole stated that while “legislation is going to be verydifficult until there is a full understanding” of where the Supreme Court would leave the state of the law, if action werenecessary his assessment was: “there has been such an avalanche of feeling about racial discrimination. You are notgoing to get any votes in this committee for racial discrimination.” Id. at 254. Some statements also suggested SenateDemocrats were not eager to enact a bill that would get the Reagan administration off the political hook, id. at 244, 247,and that might imply that the prior IRS interpretation had not been legally proper, id. at 241, 244. But there was no doubtthat their enactable preference would, if necessary, be for a statute that denied a tax exemption to racially discriminatoryschools.

233 461 U.S. at 601.234 Id. at 601-02. For social clubs, enactment was necessary to express this political preference because they

were not governed by the IRS interpretation that produced this result for schools.235 See Elhauge, Preference-Eliciting, supra note .

68

A skeptic might further worry: but if current enactable preferences really supported a denial of taxexemption, then why didn’t Congress enact this proposed bill? The answer seems plain: since the issue wasalready before the U.S. Supreme Court, Congress figured it might as well wait to see whether the Courtresolved the problem for it. Bob Dole, the Chairman of the relevant Senate committee, was quite explicitabout this, adding that: “I know we can’t suggest the Court go ahead and make that decision but hopefullythey read the papers.”232 This sort of legislative inertia is hardly unusual, and also explains why Congresshad not bothered to enact any legislation codifying the IRS interpretation in the preceding 12 years, eventhough it did enact legislation denying a tax exemption to racially discriminatory social clubs.233 Enactinganything requires the expenditure of time and political effort that will often not seem worthwhile (comparedto other pursuits), especially when the enactment will not change the status quo. The fact that Congresshas failed to enact something does not mean it would not be supported by enactable preferences ifnecessary. Thus, the Court correctly inferred that the enactment denying a tax exemption to raciallydiscriminatory social clubs did not mean Congress wanted a different rule for racially discriminatoryschools, but most likely meant that, if enactment were necessary, then enactable political preferencesfavored denying tax exemptions to racially discriminatory entities.234

Naturally, any estimate of enactable political preferences is necessarily less accurate than actuallywaiting to see what gets enacted. But the whole point of using preference-estimating default rules is tominimize political dissatisfaction for issues too minor to provoke legislative action, or in the interim beforethe legislature acts, and to free the political process from the needless burden of making enactments it wouldprobably make if time and political energy were not scarce. When what the government would enact issufficiently uncertain, a preference-eliciting default rule may well be called for to provoke legislativeresponses,235 but when official actions provide a sufficiently reliable estimate of current enactablepreferences, then those should be followed. And the necessary threshold of reliability declines the moreancient any enacting legislative preferences, as with the 1894 statute in Bob Jones.

Still, sometimes the current or subsequent preferences are quite general, whereas the enactingpreferences were quite specific. In such cases, canons of construction provide that: “Where there is noclear intention otherwise, a specific statute will not be controlled or nullified by a general one, regardless

236Morton v. Mancari, 417 U.S. 535, 550-51 (1974) (collecting cases).237 Id.238 Id. at 548-49.239 See, e.g., Square D v. Niagra Frontier Tariff Bureau, 476 U.S. 409, 424 n.34 (1986); City of Oklahoma City v.

Tuttle, 471 U.S. 808, 818 n.5 (1985); Eskridge, Overruling Statutory Precedents, GEO. L.J. 1361, 1362-63 (1988) [herinafter“Eskridge, Overruling”]; Levi, An Introduction to Legal Reasoning, 15 U. CHIC. L. REV. 501, 540 (1948).

240 Eskridge, Overriding, supra note , at 397-400.241 Id. at 399-400 & n.218 (38% of cases overruling federal statutory precedent relied on indications of current

congressional preferences).242 Id. at 399. 243 Congress overrode 121 Supreme Court statutory interpretations from 1967-1990, and 98 from 1975-1990. Id.

at 338. This amounts to 5-6 decisions a year, which is 6-8% of the 80 statutory interpretations the Court on averageissued a year. Id. at 339 n.15.

69

of the priority of enactment.”236 Under this canon, a old statute can trump a later one if the former is morespecific. Thus, the Court ruled that a general statute against employment discrimination should yield to anearlier specific statute allowing preferences for Native Americans.237 This might seem inconsistent with apolicy of maximizing current preferences, but is not. What the Court recognizes is that often enactors adoptgeneral principles without foreseeing certain specific implications, and that the government’s views on thosespecific implications can generally be ascertained more reliably by more specific legislation, even when itis earlier in time. Indeed, in that case the Court relied on the confirming evidence that, after enacting thegeneral anti-discrimination statute, the subsequent Congress enacted two new preferences for NativeAmericans, thus confirming that its general political views did not govern this specific issue.238

C. When to Overrule Statutory Precedent.When a court has itself interpreted a provision in a prior opinion, that creates statutory precedent

that might be considered binding under stare decisis. Indeed, such statutory precedent are often said toenjoy a doctrine of “super-strong” stare decisis.239 If a current preferences default rule governed, onewould expect this nominally super-strong rule to be far weaker when it conflicted with reliable evidence ofcurrent enactable preferences. Which is in fact what we see.

As Professor Eskridge has persuasively shown, statutory stare decisis tends to be followed whenthe statutory precedent conforms with current legislative preferences but not when it conflicts.240 Courtdecisions sticking to statutory precedent often rely expressly on indications that the current legislature favorsthe precedent, whereas court decisions overruling statutory precedent frequently rely expressly onindications that the current legislature disfavors that interpretation.241

Often court decisions about whether to overrule statutory precedent make no express statementabout current enactable preferences. Even then an implicit effort to conform to current enactablepreferences can be inferred from the fact that decisions overruling statutory precedent are the least likelyto be legislatively overridden. Indeed, an astonishingly low 0% of Supreme Court cases overrulingstatutory precedent were overridden by Congress from 1967-1990.242 This is far below the 6-8% rateat which Congress generally overrides statutory interpretations by the Supreme Court.243

244 See, e.g., Monell v. New York City Dept. of Social Servs., 436 U.S. 658, 690-96, 700-01 (1978).245 See supra at III.C; Boys Markets v. Retail Clerks Union, 398 U.S. 235, 244-53 (1970); Gulfstream Aerospace

v. Mayacamas, 108 S. Ct. 1133, 1138-43 (1988); Eskridge, Overruling, supra note , at 1390-91.246 See Business Electronics v. Sharp, 485 U.S. 717, 732 (1988); Easterbrook, Statutes’ Domains, supra note , at

544. See generally Eskridge, Overruling, supra note , at 1376-78. Sometimes this is mistakenly taken to mean that anystatutory language in the antitrust statutes confers the same common lawmaking authority on courts, id. at 1378-81, butit is only general language like “restraint of trade” or “monopolization” that conveys that meaning. Thus, it is notanomalous that the Court has simultaneously been careful about statutory stare decisis when invoking more technicalantitrust decisions about the jurisdictional reach of the statute, even though some notable scholars have thought so.Id. at 1378-81, 1414-21 (collecting such cases and asserting they are anomalous).

247 See Elhauge, Preference-Eliciting, supra note , at VIII. The lack of a legislative override does not alonesuffice to provide that information, but to the contrary standing alone strongly supports retention. See id.

248 See, e.g., Monell, 436 U.S. at 696-99 (relying on very general inferences drawn from current legislative action).249 See supra Section II.A (considering case where there are no reliable indicators of enacting governmental

preferences and some indication of current governmental preferences).250 591 U.S. 164 (1989).

70

None of this is to say one can establish a simple one-to-one correspondence between decisionsoverruling statutory precedents and a current preferences default rule. The connection is far morecomplicated because many cases overrule statutory precedent that also failed to comport with thepreferences of the enacting legislature. Sometimes, the Court recognizes that it incorrectly identifiedenacting legislative preferences in the statutory precedent.244 In other cases, changed circumstances maymean that the precedent no longer conforms to initial legislative preferences.245 Or perhaps the initiallegislative preference was to delegate the matter to ongoing judicial resolution, and thus supports no specialpreference for the first precedent decided as part of that common law. The classic example is judicialdevelopment of what “restraint of trade” or “monopolization” means under antitrust law.246 Such adelegation to a common law process implies that the legislative preference is to have judges correctdecisions that proved mistaken, not stick with them. Finally, on some occasions, the statutory precedentreflects a preference-eliciting default rule that should be abandoned because new information has arisenabout the likelihood it will elicit any legislative reconsideration.247 In all these cases, overruling the statutoryprecedent does not conflict with enacting legislative preferences.

When overruling statutory precedent does not conflict with enacting legislative preferences, it is notsurprising to see courts use subsequent legislative history far more liberally in estimating and followingcurrent legislative preferences.248 Since the precedent fails to provide any reliable indication of enactingpreferences, the interpretive strategy that maximizes political preference satisfaction includes somewhat lessreliable evidence of current legislative preferences.249 Indeed, when precedent conflicts with enactinglegislative preferences, legislative preferences cannot counsel against overruling that precedent unless thereis affirmative evidence that current enactable preferences favor that precedent.

But when there is affirmative evidence that current enactable preferences favor a precedent, itshould not be overruled even though the court has come to the conclusion that the precedent likely conflictswith enacting preferences. A prominent example was Patterson v. McLean Credit Union.250 There theSupreme Court admitted to some uncertainty about whether a precedent interpreting Ҥ 1981 as prohibiting

251 Id. at 174 (emphasis added).252 Id. 253 407 U.S. 258 (1972). See, e.g., Eskridge, Overruling, supra note , at 1381, 1385 (calling the decision “comical”

and “silly”).254 See Federal Baseball Club v. National League, 259 U.S. 200, 208-09 (1922).255 See, e.g., Eskridge, Overruling, supra note , at 1380-81 (collecting authorities).256 407 U.S. at 281-82 & n.18; 15 U.S.C. §§ 1291, 1294.257 See Flood v. Kuhn, 407 U.S. 258, 281-82 (1972); Eskridge, Overruling, supra note , at 1404. The Court

acknowledged this baseball exemption was inconsistent and illogical because other sports had subsequently been heldto be subject to antitrust scrutiny. 407 U.S. at 282-84. But there is no requirement that political preferences conform tojudicial notions of consistency. See supra at II. Moreover, where neither the baseball exemption, nor the other sports’nonexemption, conflicted with any affirmative indication of current enactable preferences, it was not in fact illogical to

71

racial discrimination in the making and enforcement of private contracts is right or wrong as an originalmatter.”251 Without resolving whether the precedent matched enacting preferences, the Court held it shouldnot overturn the precedent because recent enactments indicated that current congressional preferencesembodied a “deep commitment to the eradication of discrimination” consistent with that precedent.252

Where current enactable preferences either affirmatively agree or disagree with statutory precedent,applying a current preferences default rule is relatively clear. But what about when the current legislatureis split or has simply failed to act? Above I noted that mere legislative inaction is generally an insufficientbasis for concluding that legislative preferences have changed, and thus such inaction does not justifyfollowing an otherwise nonbinding interpretation that seems to conflict with enacting preferences. Butwhere there is statutory precedent by the same court that otherwise would be binding as a matter of staredecisis, the matter is different. Then legislative inaction can be relevant because it indicates the absenceof any change in enactable legislative preferences that might justify overruling the initial statutory precedent.In short, while legislative inaction cannot affirmatively establish current enactable preferences, it cansometimes rebut a claim that current enactable preferences have changed in ways that require overrulingstatutory precedent.

This seems the best explanation for the Supreme Court’s much-maligned 1972 decision in Floodv. Kuhn,253 which declined to overrule statutory precedent that for five decades had held the business ofbaseball was not “commerce” subject to antitrust scrutiny. “Commerce” is not a self-defining term, but thefirst case announcing this interpretation probably reflected accurately enacting (1890) and then-current(1922) legislative preferences about whether the antitrust laws should apply to baseball.254 The mainargument for overruling this interpretation in 1972 was that, because baseball had developed into a bigbusiness and other sports had been judicially denied any antitrust exemption, the current legislature wouldprefer to abandon the baseball antitrust exemption.255 But the Court observed that Congress hadrepeatedly rejected bills that would eliminate the baseball antitrust exemption. The only statute it enactedexpanded this exemption to allow other sports to collectively bargain about television rates, and expresslystated that it did not intend to otherwise change any existing antitrust exemption or nonexemption for anysport.256 Likewise, the only other bills to get out of committee (one passed by each legislative chamberat different times) would have expanded the baseball exemption by extending it to other sports.257 The

differentiate the sports based on their reliance on different precedent , which is what the court did, see 407 U.S. at 274,283. While such reliance interests should not trump the goal of political satisfaction, when political preferences fail toreliability cut in either direction, it does advance that goal to generally have interpretations stable because that inducesmore behavioral reliance. See Elhauge, Preference-Eliciting, supra note , at XI.B.

258 See, e.g., Eskridge, Overruling, supra note , at 1404-06. 259 Id. at 274, 283. Further, it can be shown that even if the initial adoption of the baseball exemption reflected

a preference-eliciting default rule, the lack of a legislative override justifies retaining that precedent absent any otherdemonstrable change in legislative conditions. See Elhauge, Preference-Eliciting, supra note , at VIII.

72

standard critique is that such failures to act need not indicate congressional approval of the baseballexemption — it may merely mean Congress was too busy or divided to take any action.258 But the Courtwas not relying on legislative inaction to support an interpretation that was initially flawed. The Court wasinvoking this legislative inaction to rebut the argument that changed political preferences meant it should nowabandon otherwise binding precedent that did seem to accurately reflect enacting legislative preferences.For such a rebuttal, it sufficed that there were no indication that enactable preferences had changed, andthat to the extent current political preferences could be ascertained they cut against changing theinterpretation. Thus, this was an appropriate case to insist, as the court did, that any override be providedby affirmative legislative action.259

Apparent inconsistencies in judicial willingness to follow statutory stare decisis or pay attention tosubsequent legislative inaction have been a favorite target of academic critique. But default rule analysisindicates that this critique cannot justifiably be founded (as it generally is) merely on the fact that thestatutory stare decisis is sometimes applied and sometimes is not, and that subsequent legislative inactionis sometimes deemed relevant and sometimes is not. Whether to apply statutory stare decisis often should(and does) turn on whether the statutory precedent conflicts with reliable indicators of current enactablepreferences. Likewise, subsequent legislative inaction generally should not (and is not) used to justifydeviating from statutory precedent that conforms to enacting preferences, but should (and is) used to rebutan argument that enactable preferences have changed in ways that justify such a deviation.

D. Evidence That Legislatures Prefer a Current Preferences Default RuleThe analysis so far is that it makes sense for legislatures to prefer a current preferences default rule,

and that the actual judicial practice has been to use one when there is reliable official evidence of currentenactable preferences. But is there any direct evidence that legislatures actually prefer a currentpreferences default rule? Not much, but what there is of it is confirmatory. Since this is the actual defaultrule, one would think legislatures would opt out of it if it did not maximize political satisfaction. In fact, nolegislature in the United States has done so even though each has enacted a code of statutory construction.

This evidence standing alone is weak because many of these codes cover only relativelyunimportant matters and even the legislatures that enacted more substantive construction codes may havenever considered this issue. But the only state legislatures that have adverted to the issue have adoptedelements of a current preferences default rule. Two state legislatures have directed that: “When the wordsof a law are not explicit, the intention of the legislature may be ascertained by considering . . . Legislative

260 M INN. STAT . § 645.16; 1 PA. CON. STAT . §1921.261 N.Y. MCKINNEY'S STAT . §75.262 M INN. STAT . § 645.17; 1 PA. CON. STAT . §1922.263 COL. REV. STAT . §§ 2-4-203; IOWA CODE § 4.6; M INN. STAT . § 645.16; N.Y. MCKINNEY 'S STAT . §129; NEW

MEXICO STAT ANN, 1978, § 12-2A-20; NORTH DAKOTA CENTURY CODE §1-02-39; 1 PA. CON. STAT . §1921; OHIO REV. CODE

§ 1.49; TEXAS GOV’T CODE 311.023.264 NEW MEXICO STAT ANN, 1978, § 12-2A.265 Chevron U.S.A. Inc. v. Natural Resources Defense Council, Inc., 467 U.S. 837 (1984). 266 Merrill & Hickman, Chevron’s Domain, 89 GEO. L.J. 833, 833 (2001). Some areas, like labor and tax law, also

had longstanding deference doctrines. Id. at 838-39.

73

. . . interpretations of the statute.”260 Similarly, another state legislature has provided that: “The constructionof a statute by the Legislature, as indicated by the language of later enactments, is entitled to considerationas an aid in the construction of the statute, but is not generally regarded as controlling.”261 The fact that itis not controlling emphasizes that this is only a default rule, not a ground for changing the actual meaningof the statutory language. Two state legislatures have also adopted the default rule in favor of prevailinginterpretations that the legislature elected not to alter when it enacted “subsequent laws on the same subjectmatter”262

Finally, many state legislatures have also adopted the interpretive rule that, when a statute isambiguous, a court may consider the administrative construction of the statute.263 I could find no statelegislature that rejected this default rule. One state legislature even adds that the court can consider anagency’s interpretation of a “similar statute.”264 As we will see next, this rule of deferring to agencyinterpretations can also be explained as a current preferences default rule.

VII. DEFERENCE TO INTERPRETATIONS BY

POLITICALLY ACCOUNTABLE AGENCIES

In all the above cases, courts are tracking political preferences directly evidenced by the mostrecent legislature and executive. But sometimes, especially when there is no indication from the legislatureor executive themselves, the next best indicator of current enactable preferences are the acts of a politicallyaccountable agency. The U.S. Chevron doctrine can thus be understood as a current preferences defaultrule.265 Indeed, it turns out that this theory better explains Chevron’s justification and limited scope thanprevailing theories.

A. Chevron as a Current Preferences Default RuleUnder Chevron, U.S. courts defer to agency interpretations of ambiguous statutes. Agency

deference was not new, but prior to Chevron, courts generally deferred only if Congress had expresslydelegated interpretive authority to the agency.266 Chevron changed the general default rule so that, unlessCongress indicates otherwise, courts must defer to interpretations by the agency charged with

267 Id. at 833.268 See Wayne Washington, Ashcroft Supports Affirmative Action Rule -- Files a Brief Backing Policy in Colo.

Case, BOSTON GLOBE A3 (Aug. 11, 2001)269 On the traditional and new presidential mechanisms for controlling agencies, see Kagan, Presidential

Administration, 114 HARV. L. REV. 2245, 2274-2319 (2001). On the power of legislative oversight and budgetary review,see, e.g., JOEL D. ABERBACH, KEEPING A WATCHFUL EYE (1990); Dwyer, The Pathology of Symbolic Legislation, 17ECOLOGY L.Q. 233, 245-46 (1990); Barry R. Weingast & William J. Marshall, The Industrial Organization of Congress,96 J. Pol. Econ. 132 (1988); McNollgast, Administrative Procedures , supra note ; Morris P. Fiorina, CongressionalControl of the Bureaucracy: A Mismatch of Incentives and Capabilities, in CONGRESS RECONSIDERED 332 (Lawrence C.Dodd & Bruce I. Oppenheimer eds., 2d ed. 1981).

270 The personal political views of agency heads may be a different matter, but what matters here are the revealedpolitical preferences implied by what they actually do.

271 See McNollgast, The Political Origins of the Administrative Procedure Act, 15 J.L. ECON. ORG. 180, 181, 185-86, 198-99 (1999) [hereinafter McNollgast, APA] (noting that procedures force agencies to response to policy preferencesof groups other than those favored by the President and summarizing “fire alarm” theory of congressional oversight);Kagan, supra note , at 2258 (same).

272 See infra at VII.C.

74

administering the statute.267

As a hermeneutic matter, this choice of a general default rule would be odd. Agencies may be veryfamiliar with the statute in question, but they are not particularly skilled in legal interpretation, at least nomore skilled than courts. Further, agencies have certain biases (such as a bias in favor of expanding theirpower) that might distort their interpretation. Courts thus seem better hermeneutic interpreters of statutorymeaning.

But Chevron does make sense as a current preferences default rule, to cope with cases leftunresolved by hermeneutic theory. In the United States, agency heads have been nominated by thePresident and confirmed by the Senate based largely on the acceptability of their policy views in a specificpolicy area. Executive interviews and Senate confirmation hearings pin down nominees on the issues likelyto arise during what will be a relatively brief tenure. Consider the close questioning of John Ashcroft duringhis confirmation hearings, which produced commitments that help explain his later decision as AttorneyGeneral to defend the constitutionality of an affirmative action plan he had opposed as a Senator.268

Agency heads are also supervised by the executive, and subject to extensive committee oversightand budgetary review in the legislature.269 They are normally removable by the executive, and even inindependent agencies, they serve limited terms and thus depend on the executive and legislature for theirre-appointment. Agency heads also depend on the executive and legislature for cooperation on futureinitiatives, including any new legislation, executive action, or appointment of subordinates. The policy viewsthat govern the actions of agency heads thus generally come about as close to an accurate barometer ofcurrent political preferences as courts can get.270 This is particularly true for agency action taken aftercurrent political forces have had an opportunity to participate. Such participation not only helps thoseactors directly inform agencies about prevailing political preferences, but also enables them to alertCongress when a problematic issue has arisen.271 Indeed, as we will see, an agency that acts without suchan opportunity for political participation does not get Chevron deference.272 Agencies that get Chevron

273 See Kagan, supra note , at 2254-60, 2264-69, 2274-2319 (observing while different theories emphasizePresidential, Congressional and interest group influence respectively, the combined literature indicates that all of themhave real influence on agency action).

274 See, e.g., Kagan, supra note , at 2259, 2272-74 (noting limits of Congressional and Presidential influence onagencies); Cynthia R. Farina, Statutory Interpretation and the Balance of Power in the Administrative State, 89 COLUM.L. REV. 452, 502-03 (1989) (same); M cNollgast, Positive Canons: The Role of Legislative Bargains in StatutoryInterpretation, 80 GEO. L. J. 705, 737 (1992). See generally McNollgast, APA, supra note , at 184-86 (summarizing problemof agency drift and describing strategies to restrain it).

275 See U.S. CONSTITUTION . 276 Even if Congress wanted to lower judicial budgets in response to adverse judicial decisions, this would be

a poor tool because there are so many judges. Congress would thus have to punish all judges to punish one, and nosingle judge would have much incentive to render cooperative decisions to help judges collectively protect theirbudgets. See Elhauge, Interest Group Theory, supra note , at 85-86.

277 In contrast, the Administrative Procedure Act does not bar ex parte contacts between agency personnel andprivate parties in notice-and-comment rulemaking. See Kagan, supra note , at 2280 n.142.

278 It is most revealing when it consists of actual legislative action, but least revealing when it consists ofstatements by some of many legislators. See supra Part III. A related advantage of agency deference is that it providesa single source for the best estimate of current preferences, rather than a multiplicity of possible conflicting sources.

75

deference are thus subject to a combination of political influences by the executive, legislature, and outsidepolitical groups.273 Such agency action generally provides the best available estimate of where currentenactable preferences lie.

True, agency accountability to political oversight is far from perfect, in part because such oversightprobably gives the President and oversight committees somewhat greater power vis-a-vis other legislativeparticipants than they normally possess in securing enactments.274 Thus some deviation between agencyaction and current enactable preferences must be expected. But the question is not whether agenciesperfectly reflect current political preferences. The question is whether agencies generally reflect currentenactable preferences better than judicial guesses about where those preferences lie. If they do, thenpolitical satisfaction can be maximized by a general default rule of deferring to agency interpretations.

Compared to agency heads, judges were appointed a long time ago, with far less inquiry into theirpolicy views, whose relevance is limited neither to a specific policy area nor a short period of time thatmakes most relevant issues foreseeable. Judges are also politically insulated. They serve life terms, arenot removable and have salaries that cannot be lowered.275 They depend little on budgetary support, whichanyway as a matter of practice is unaffected by their individual decisions.276 Perhaps more important, theyare insulated by judicial norms and ethical canons from general contact with, and information flow from,political officials and groups.277 Judges can thus generally ascertain current political preferences only froma cold record of legislative action. Absent the sort of recent legislative action described in Part III, sucha record can be quite unrevealing about where current enactable preferences lie.278

The details may differ in other nations, but the general point likely remains valid. Being moreaccountable and informed by current political preferences than the judiciary, agency action is more likelyto accurately reflect current enactable preferences than judicial guesses about what current political forceswould enact. Indeed, agency responsiveness to enactable preferences should even be stronger in

279 Cf. Kagan, supra note , at 2347-48 (noting evidence that independent agencies, which are less likely to beprotected from overriding legislation by Presidential veto, are more responsive to Congress).

280467 U.S. at 842-43. 281Id. at 865-66.282 Id. at 865-66.

76

parliamentary systems where the agency does not enjoy the protection from legislative overruling sometimesprovided by deadlock between the President and Congress.279 By following agency interpretations toresolve statutory indeterminacies when the record leaves statutory meaning unknown, courts thus followa default rule that is best calculated to minimize current political dissatisfaction.

The language of Chevron tracks this paper’s distinction between hermeneutic matters (on whichthere is no deference) and the policy choice of how to resolve the indeterminancy (on which there isdeference). When courts can divine the meaning of a statute, Chevron provides that this meaning must befollowed notwithstanding contrary agency interpretations.280 On the other hand:

When a challenge to an agency construction of a statutory provision, fairly conceptualized, reallycenters on the wisdom of the agency's policy, rather than whether it is a reasonable choice withina gap left open by Congress, the challenge must fail. In such a case, federal judges--who have noconstituency-- have a duty to respect legitimate policy choices made by those who do. . . . Whileagencies are not directly accountable to the people, the Chief Executive is, and it is entirelyappropriate for this political branch of the Government to make such policy choices--resolving thecompeting interests which Congress itself either inadvertently did not resolve, or intentionally leftto be resolved by the agency charged with the administration of the statute in light of everydayrealities.281

This language justifies the Chevron default rule because of its ability to track current political preferencesby following the interpretations of agencies that were best positioned to ascertain where the current balanceof political interests lay. The Chevron Court squarely considered the alternative default rule – that suchmatters should be left to judicial judgment -- and rejected that rule as insufficiently responsive to currentpolitical preferences.

Judges are not experts in the field, and are not part of either political branch of the Government.Courts must, in some cases, reconcile competing political interests, but not on the basis of thejudges' personal policy preferences. In contrast, an agency to which Congress has delegatedpolicy-making responsibilities may, within the limits of that delegation, properly rely upon theincumbent administration's views of wise policy to inform its judgments. . . . The responsibilities forassessing the wisdom of such policy choices and resolving the struggle between competing viewsof the public interest are not judicial ones: "Our Constitution vests such responsibilities in thepolitical branches."282

The Chevron Court also made clear that the proper default rule was dynamic. The Court chastisedthe appellate court for adopting “a static judicial definition” of the statutory term “when it had decided thatCongress itself had not commanded that definition,” and rejected the argument that deference to agency

283 Id. at 842, 863-64.284 Id. at 842-43.285 See supra Part II.286 Merrill & Hickman, supra note , at 863.287 Id. at 864-65 (summarizing the literature). See also Manning, Textualism as Non-Delegation, supra note ,

at 706-14 (also offering such an argument).288 Id. at 866-67.289 Id. at 870-71.290 Id. at 871; Antonin Scalia, Judicial Deference to Administrative Interpretations of Law, 1989 DUKE L.J. 511,

517. See also Farina, supra note , at 468-76 (detailing the weaknesses of basing Chevron on legislative intent).291 Merrill & Hickman, supra note , at 871.

77

interpretations should not apply where, as here, the agency had changed its interpretation over time.283

Again, the distinction between hermeneutic and default rule analysis helps understand the point. Changingan interpretation over time does undermine the credibility of a hermeneutic judgment regarding statutorymeaning, for it suggests the agency itself is inconsistent or uncertain about what that meaning is. Butchanging an interpretation does not undermine a default rule judgment about where current politicalpreferences lie, for an accurate reading of those would change over time.

Also consistent with this article’s analysis, the Court rejected dynamic statutory interpretation (byagencies or courts) in cases where the enacting legislature’s meaning was clear.284 It thus rejected theproposition, advanced by some scholars, 285 that changing political preferences should change statutorymeaning itself. Instead, this current preference default rule is limited, as is default rule analysis generally,to cases where statutory meaning is unclear.

B. Alternative Theories of ChevronA rich literature suggests numerous alternative theories of Chevron. One topic concerns what

source of law underpins the Chevron doctrine. As Thomas Merrill and Kristin Hickman summarize in theirrecent tour-de-force on Chevron’s domain, the current literature identifies three possible sources.286 Oneline of articles argues that Chevron rests on constitutional separation of powers and anti-delegationnorms.287 But Chevron does not cite any constitutional provision or rest on any constitutional argument,and (if anything) seems to enhance rather than deter delegation of legislative responsibilities to agencies.288

Thus, few scholars are convinced that the source of law for Chevron is constitutional.A second theory, to which most subscribe, is that when Congress enacts an ambiguous statute and

is silent about whom it wants to resolve that ambiguity, it implicitly intends to delegate interpretive authorityto the administering agency.289 The problem is that even adherents to this theory – like Merrill andHickman, and Chevron’s leading judicial champion, Justice Scalia – acknowledge that the evidence of suchenacting congressional intent is “weak” and even “fictional.”290 At the time Chevron was decided,Congress had no reason to think that enacting a vague statute delegated interpretation to agencies ratherthan courts, and had never indicated any disapproval of the less deferential judicial interpretation that thenprevailed.291 Moreover, Chevron itself acknowledged that statutory ambiguity might be created

292 467 U.S. at 865.293 5 U.S.C. §706 (2000); Anthony, The Supreme Court and the APA: Sometimes They Just Don't Get It, 10

ADMIN. L.J. AM. U. 1, 9-11 (1996). 294 McNollgast, APA, supra note , at 189-192, 200, 203-04, 213-215.295 Id.296 See supra Part II.297 Merrill & Hickman, supra note , at 687-68 & n.191 (collecting sources). 298 Id. at 868-70 & n.197. Merrill and Hickman also worry that this theory is hard to square with the APA, but

ultimately concede that understanding Chevron as a canon of construction avoids this problem. Id. at 868.

78

“inadvertently” because Congress “simply did not consider the question.”292 Nor had Congress given anyother affirmative indication of intent to delegate interpretive decisions to agencies.

To the contrary, the most relevant statute Congress enacted, the Administrative Procedure Act,provides an opposite indication by stating that courts “shall decide all relevant questions of law [and]interpret constitutional and statutory provisions”when they review agency action.293 Further, the politicalanalysis of McNollgast shows the enactment of the Administrative Procedure Act was motivated by acongressional desire to constrain agency discretion.294 Thus, it seems dubious to conclude that theenactment of a vague statute itself indicates a congressional intent to give agencies discretion over itsinterpretation. But seeing the issue as a current preferences default rule makes more sense of Chevron’simplicit interpretation of the Administrative Procedure Act. For on McNollgast’s analysis, what the NewDeal Congress did was enjoy the benefits of agency discretion when it was in power and then, when it sawits dominance ending, tried to lock in its political gains by constraining that discretion in the future.295 Thisis a special case of the general situation (described above) when a legislature enjoys the benefits of acurrent preferences default rule, then at the end of its reign tries to enact an interpretive rule adopting anenacting preferences default rule for future cases.296 My prediction that courts would treat such an opt-outwith hostility is consistent with the narrow reading given the Administrative Procedure Act by the Chevrondoctrine, and provides a normative justification for that otherwise anomalous reading.

A third theory is that Chevron is a species of federal common law, a judicially devised canon ofinterpretation.297 Merrill and Hickman argue against this theory on the grounds that it: (1) cannot explainwhy Chevron is mandatory, (2) does not explain how to prioritize the Chevron canon against othercanons, (3) fits poorly with the normal understanding that canons are used to divine statutory meaning, (4)does not match the language in court opinions that refers to implicit congressional desires, and (5) robsChevron of its normative force.298

Understanding Chevron as default rule, however, provides a fourth source of law that avoids theweaknesses of the second and third approach. Like the second theory, a default rule approach rests ona view about what Congress wants. But it does not rely on any strained conclusion that the enactingCongress “intended” any delegation to agencies by failing to cure all ambiguities in its statute. Rather, itsjustification is that, where the enacting government has left no clear intent, it would want courts to choosethe default rules that minimize political dissatisfaction, and that Chevron is such a rule. Nor is theAdministrative Procedure Act a problem because the “relevant questions of law” that courts must resolve

299 See also Elhauge, Preference-Eliciting, supra note (describing how many canons are instead designed toelicit more precise indications of political preferences)..

300 See Peter L. Strauss, One-Hundred-Fifty Cases Per Year: Some Implications of the Supreme Court's LimitedResources for Judicial Review of Agency Action, 87 COLUM. L. REV. 1093, 1118-29 (1987). At the time it was 150 casesa year, now it is about 75, and most of these do not involve reviews of agency interpretations.

301 Under this practice of nonacquiescence, the agency declines to follow an appellate interpretation in othercircuits or even to other persons in the same circuit who were not party to the appeal. Id. at 1110-1113.

302 See Elhauge, Preference-Eliciting, supra note , at XI.C (considering this alternative default rule).

79

to “interpret . . . statutory provisions” include deciding which default rule to choose when statutes areambiguous, so that when courts choose Chevron they are deciding the relevant legal question of statutoryinterpretation.

Like the third theory, the default rule approach is judicially created. But because it is created basedon a theory of what Congress would want if it thought about the general default rule issue, it is broadlyconsistent with the language in court opinions that refers to implicit congressional desires. This also explainswhy Chevron is mandatory and has such normative force: it conforms to the ultimate aim in statutoryinterpretation of maximizing the satisfaction of enactable political preferences. It does fit poorly with thenotion that statutory canons are exclusively about divining statutory meaning, but fits well with this paper’sthesis that instead these canons can generally be better understood as default rules calculated to maximizepolitical satisfaction when statutory meaning is ambiguous.299 This theory also provides an overarchingstructure and objective to prioritize what otherwise seems like a melange of conflicting canons.

Other theories focus less on explaining the source of law for Chevron than on establishing a goodpolicy justification for it. One theory, well-developed by Peter Strauss, is that Chevron deference furthersimportant interests of uniformity in a administrative system where diverging lower court interpretationswould otherwise be the norm because the U.S. Supreme Court decides relatively few cases each year.300

This is certainly a important virtue of agency deference, and does ameliorate the problems resulting fromthe agency practice of sometimes declining to follow circuit interpretations in other cases until the agencyhad secured review in the Supreme Court.301 But the fit between uniformity and Chevron is not great.Uniformity does not explain why, as discussed in Section C below, the Supreme Court does not giveChevron deference to many sorts of agency interpretations that would confer the necessary uniformity butlacked sufficiently reliable opportunities for political participation. Nor does uniformity explain why theU.S. Supreme Court itself defers to agency interpretations, or why there should be agency deference in lessfractured state or foreign appellate systems. Uniformity does not justify deference to agency decisions thatare either formally or de facto appealable only to the one intermediate appellate court (like the FederalCircuit or the D.C. Circuit), which can thus create national uniformity in the appellate system. Further,uniformity cannot explain why the appellate courts do not also defer on whether a statute has anunambiguous meaning and what that meaning is, issues on which an agency decision can also cure thedisuniformity that might otherwise result from conflicting circuit decisions. Finally, if uniformity were thegoal, then it could equally be advanced by the alternative default rule of interpreting all statutory ambiguitiesto mean the statute has the most narrow regulatory effect possible.302 We need some theory to explain why

303 See supra at VI.D.304 Richard B. Stewart, The Reformation of American Administrative Law, 88 HARV. L. REV. 1667, 1683 (1975).

See also Farina, supra note , at 524 (collecting sources).305 See generally Scott Brewer, Scientific Expert Testimony and Intellectual Due Process, 107 YALE. L.J. 1535

(1998) (discussing how this problem exists even for scientific experts in courts of law).

80

the uniformity-promoting default rule chosen was instead to defer to agencies. The structure of the doctrineleaves the definite sense that while greater national uniformity in the U.S. appellate system is an importantbenefit of Chevron, it is really a nice side-benefit rather than the driving force behind the doctrine.

This points to another advantage to a current preferences default rule approach. It frees Chevronfrom just being a parochial theory limited to the particular case of the United States. The theories abovedepend to varying degrees on peculiarities of the U.S.’s history, constitutional or statutory provisions,agency structure, appellate system, or practice of judicial common law. They thus render Chevron adoctrine to some degree limited to a particular nation under particular legal circumstances. The analysishere, in contrast, provides a more general justification for judicial deference to statutory interpretations bypolitically accountable agencies, a justification that would extend to legislation enacted elsewhere in theworld or by political subunits of the United States. For example, given the fact that many state legislatureshave adopted (and to my knowledge none has rejected) the Chevron rule,303 a satisfying theory forChevron should not rest on peculiarities of the limited docket of the U.S. Supreme Court and its inabilityto police circuit conflicts in statutory interpretation.

The leading alternative for this more general justification for Chevron is that agencies have greaterpolicy expertise than courts. But the legal realists’ hope that legal ambiguities could be resolved byobjective policy expertise has long ago grown quaint. Although expertise can theoretically informdecisionmakers about the likely consequences of a given interpretation, even in theory it cannot resolvewhich statutory interpretation has the “best” policy implications. Such ultimate resolution depends onpolitical preferences about how to weigh those consequences. Both theory and experience has “sappedfaith in the existence of an objective basis for social choice.”304 In practice, it is rare to find the field ofsocial policy where there are not two experts on each side of an issue, each retained by an opposing camp,undermining any claim to an objective expert resolution. This is true even for many scientific or empiricalquestions.305 Indeed, on probably most issues the two U.S. political parties are divided less by differingnormative assessments than by contrary views (informed by rival expert academics and think tanks) aboutwhat empirical consequences will flow from certain policy choices. For example, rival empirical viewsseem to largely dominate policy debate on such contemporary issues as rent control, free trade agreements,global warming treaties, ballistic missile defense, the patients’ bill of rights, Medicare reform, tax cuts, orthe creation of individual Social Security accounts. Further, many actual agency resolutions of ambiguitiesdepend less on policy than on linguistic or hermeneutic claims at which they are probably less expert thancourts. Finally, even if objective policy expertise could resolve the relevant issue, agencies may not use it.For agencies are in fact subject to external political influences, or internal desires to expand their power or

306 See Elhauge, Interest Group Theory, supra note , at 42 (collecting literature on agency capture); JohnShepard Wiley, Jr., A Capture Theory of Antitrust Federalism, 99 HARV. L. REV. 713, 723-28 (1986) (same); WILLIAM A.NISKANEN, JR., BUREAUCRACY AND REPRESENTATIVE GOVERNMENT 36-42 (1971) (arguing that bureaucrats try to maximizetheir budget and power); Kagan, supra note , at 2354 (instead of applying expertise, agencies might respond to “interestgroup lobbying, bureaucratic (but nonexpertise-based) policy views, or bureaucratic protection of turf or other self-interest”); Stewart, supra note , at 1682-83 (“To the extent that belief in an objective ‘public interest’ remains, theagencies are accused of subverting it in favor of the private interests of regulated and client firms.”).

307 Stewart, supra note , at 1683. 308 Id. at 1684.309 See, e.g., Cass R. Sunstein, Law and Administration After Chevron, 90 COLUM. L. REV. 2071, 2088-90, 2094-95

(1990); STEPHEN BREYER, BREAKING THE VICIOUS CIRCLE: TOWARD EFFECTIVE RISK REGULATION 61 (1993). 310 See United States v. Mead Corporation, 121 S. Ct. 2164, 2171-72, 2175-76 (2001) (summarizing factors under

Skidmore deference). Skidmore is the backstop doctrine that applies when Chevron deference is unavailing. Id. at 2168,2175-76.

311 467 U.S. at 865. See also id. (noting that the particular statutory scheme was “technical and complex.”)312 Id. at 865-66, quoted above in text. Accord Merrill & Hickman, supra note , at 855 (agreeing that while

Skidmore is based on agency expertise, that plays little if any role under Chevron).313 See supra Section A Agency uniformity is somewhat different in that, while it is not a sufficient condition

for Chevron deference, it may be a necessary condition. See infra Section C.

81

express their own ideological preferences, any of which might distort expert judgment.306 In short, rather than reflecting objective expertise, “the exercise of agency discretion is inevitably

seen as the essentially legislative process of adjusting the competing claims of various private interestsaffected by agency policy.”307 This is troubling under an expertise model. But under a current preferencesdefault rule, the fact that the “required balancing of policies is an inherently discretionary, ultimately politicalprocedure,”308 does not undermine Chevron deference, it justifies it.

None of this is to say that agency expertise and uniformity is irrelevant. Agencies have substantiveexpertise that helps them resolve how any given political preferences should apply to particular facts, andare also better placed to coordinate, prioritize, or trade off conflicting regulatory preferences compared tocourts that see only one statutory provision or fail to see how regulating one problem can worsen others.309

But such factors are properly taken into account under the weaker deference doctrine of Skidmore, whichexplicitly depends on the degree of agency expertise, care, persuasiveness, consistency over time, anduniformity across the agency.310 All these are logical features of a doctrine based on policy expertise, forone would expect judgments based on objective expertise to be made carefully and persuasively, and tobe consistent over time and uniform across the agency (absent changed circumstances that should beevident from the record). But these features have little to do with the actual Chevron doctrine. WhileChevron does refer to the fact that “[j]udges are not experts in the field,”311 its analysis ultimately rests onthe notion that agencies will rely on political determinations of wise policy and how best to trade offconflicting interests.312 Unlike Skidmore, Chevron was untroubled by (indeed almost celebrated) the factthat the agency position changed over time with changing political preferences rather than sticking to anobjectively expert opinion.313 Unlike Skidmore, Chevron deference does not depend on any showing ofagency expertise, or the care or persuasiveness of agency decisions. Instead, as shown in the next section,

314 A different issue is raised when Congress expressly delegates agency authority to interpret a statute withoutany political input. Such agency action does not fit within the default rule provided by Chevron, but Congress has inthese circumstances opted out of that default rule, perhaps because it finds the expertise model particularly attractivefor that agency.

315 See, e.g., EEOC v. Arabian American Oil Co. (Aramco), 499 U.S. 244, 256-57 (1991); Bowen v. GeorgetownUniv. Hospital, 488 U.S. 204, 212 (1988) (“We have never applied [Chevron deference] to agency litigating positions thatare wholly unsupported by regulations, rulings, or administrative practice.”); Pub. Employees Ret. Sys. v. Betts, 492 U.S.158, 174-75 (1989) (denying deference to statutory interpretation reflected only in the agency’s brief); Gregory v.Ashcroft, 501 U.S. 452, 485 n.3 (1991) (White, J., concurring) (denying deference to EEOC’s litigation position). Incontrast, the courts do give Seminole Rock deference to agency interpretations of their own regulations even thoughthose interpretations are only advanced in litigation briefing. See Auer v. Robbins, 519 U.S. 452, 462 (1997). Althoughto Justice Scalia Auer indicated a change in the law on Chevron deference, the Supreme Court rejected that view in Mead,121 S. Ct. at 2177 n.19 (denying Chevron deference to agency statutory interpretation that was reflected in litigationbrief).

316 See Smiley v. Citibank (S.D.), N.A., 517 U.S. 735, 741 (1996).317 See infra at __ (discussing this more fully).

82

Chevron deference depends mainly on the opportunities for political participation in the relevant agencydecision. Without such an opportunity, Chevron deference is denied even though the agency decision isjust as expert and uniform as decisions that get Chevron deference.314 This would poorly fit a doctrine thattracks agency expertise or uniformity, but makes sense if the doctrine instead tracks current politicalpreferences.

C. Explaining Chevron’s LimitsThe logic of current preferences default rules also suggests limits on deference to agency

interpretations. These limits help explain why the Court defers to some agency interpretations and notothers -- an inconsistency that has often baffled commentators. Understanding Chevron as a currentpreferences default rule explains these limits better than the alternative theories.

One limit is that courts do not defer to agency interpretations that just reflect positions taken inlitigation briefs, on the ground that they were not arrived at in the normal course of its administration orexercise of rule-making powers.315 This caselaw would not make much sense if agency expertise were thebasis for deference, for an agency can provide just as much expertise and reasoned decisionmaking duringlitigation as outside of it. Nor does it fit well with a concern about uniformity, since top-level agencydecisions to take a position in litigation could confer just as much uniformity as any other top-level agencydecision. But this distinction does make sense if the real ground is deference to current politicalpreferences. Unlike positions taken in litigation briefs, agency proceedings that allow regular opportunitiesfor legislative oversight and political participation provide grounds for some reasonable confidence that theagency position reflects current political preferences. This is why the Supreme Court does defer when, inresponse to litigation, the agency promulgates a rule though notice-and-comment rulemaking.316 Althoughthis too reflects a litigation position, such a rulemaking process affords an opportunity for political input thatmakes it a more reliable indicator of current preferences.317

Similarly, the courts have never adopted the position, cleverly argued by Dan Kahan, that courts

318 See Dan M. Kahan, Is Chevron Relevant to Federal Criminal Law?, 110 HARV. L. REV. 469 (1996).319 Id. at 495-96, 518-20.320 Id. at 490; Merrill & Hickman, supra note , at 879. Outside of criminal law, Department of Justice expertness

and uniformity instead justifies weaker Skidmore deference. Kahan, supra, at 509. Inside criminal law, however, thereare grounds for a preference-eliciting default rule that explains why, rather than deferring to the Department of Justice,the courts instead apply a rule of lenity that leans against it. See Elhauge, Preference-Eliciting, supra note , at III.

321 Mead, 121 S. Ct. at 2175, 2176 n.17; Christensen v. Harris County, 529 U.S. 576, 587 (2000). Nor wouldChevron deference be triggered by more informal statements by agency officials in speeches, press releases, or pressconferences.

322 Merrill & Hickman, supra note , at 886 & n.262 (collecting cases showing that such statements, manuals andguidelines do not require notice and comment).

323 Mead, 121 S. Ct. at 2168, 2172-73; Christensen, 529 U.S. at 587.324 Mead, 121 S. Ct. at 2171-73; Christensen, 529 U.S. at 587. 325 See Merrill & Hickman, supra note , at 881.

83

should give Chevron deference to Department of Justice interpretations of criminal law.318 TheDepartment of Justice is undoubtedly expert, and can render nationally uniform interpretations throughenforcement manuals or other means in advance of undertaking any litigation.319 Such decisions thus neednot fall within the rule against deferring to agency interpretations taken during litigation, and deference wouldseem warranted if uniformity or expertise were the true basis of for Chevron. However, far from renderingitself accountable to prevailing political preferences, the Department of Justice proudly takes great painsto insulate itself from political interference, especially with respect to criminal prosecutions. There isaccordingly no reason to think that Department of Justice interpretations bear any relation to currentpolitical preferences. Thus they do not fall within a current preferences default rule, which explains whythey do not in fact get Chevron deference even though the uniformity and expertise rationales would havesuggested otherwise.320

More generally, the U.S. Supreme Court has recently clarified that it will not grant Chevrondeference to agency interpretations contained in agency letter rulings, opinion letters, policy statements,agency manuals, and enforcement guidelines.321 Again, from an expertise or uniformity perspective, thisis puzzling since such agency statements might well reflect an expert judgment applied uniformly throughoutthe agency. However, these sorts of agency statements can be made without allowing for notice to, andparticipation by, interested parties, which itself would likely trigger more serious legislative oversight andexecutive involvement.322 They thus do not ordinarily provide a sufficiently reliable assurance that theagency decision reflects prevailing political preferences.

The Supreme Court has arrived at a different formulation to reconcile its caselaw: to get Chevrondeference, agency action must have the “force of law.”323 This label is hardly self-defining. Mead held thatagency action qualifies if it involves notice-and-comment rulemaking, formal adjudication, or “other” caseswhere agencies speak with the force of law, and cites only one example (discussed below) for this residual“other” category.324 This seems hard to square with the ordinary formalistic understanding, ablychampioned by Merrill and Hickman, that actions have the “force of law” if, of their own force, they cancommand behavior or impose penalties.325 True, this formalistic understanding is consistent with the lack

326 Id. It also generally explains Chevron’s inapplicability to agencies who perform only funding, investigationor prosecution. Id. at 837.

327 See Merrill & Hickman, supra note , at 846-47, 890-92.328 Mead, 121 S. Ct. at 2172-73; Christensen, 529 U.S. at 587. 329 See infra at __ (discussing other indicia that this example fit a current preferences default rule).330 Mead, 121 S. Ct. at 2173. 331 Mead, 121 S. Ct. at 2169, 2174-74.332 Christensen, 529 U.S. at 587 (quoting Reno v. Koray, 515 U.S. 50, 61 (1995)).333 See Merrill & Hickman, supra note , at 838 n.23, 890-92 (collecting cases contrary to their position).334 See 5 U.S.C. §§556-57 (2000) (providing requirements for formal adjudication); Merrill & Hickman, supra note

, at 885-86. 335 See Envirocare of Utah, Inc. v. NRC, 194 F.3d 72, 79 (D.C. Cir. 1999), apparently overruling the cases collected

in Kagan, supra note , at 2271 n.72.

84

of deference given agency positions taken during litigation, Department of Justice manuals, or opinion lettersand the like, each of which normally require the agency to go to court in order to have any coercive effecton private parties.326 But, as Merrill and Hickman themselves argue, their formalistic force of law notionwould also extend deference to many forms of rulemaking undertaken without notice-and-comment, anddeny deference to formal agency adjudications that are not self-executing.327 Yet the Court shows no signof doing so. It continues to define the rulemaking that has force of law as “notice-and-commentrulemaking” and the adjudication that has the force of law as “formal” adjudication.328 True, Mead leftopen a small residual category illustrated by one case that involved informal rulemaking,329 but it made clearthat the most significant factor was the existence of a “notice-and-comment” procedure.330 In justifying itsactual holding denying deference, Mead repeatedly stressed the lack of any broad-based notice-and-comment opportunity.331 Likewise, Christensen denied deference based in part on the proposition thatan agency guideline that is not subject to “public notice and comment” is entitled only to “somedeference.”332 And the Court has repeatedly deferred to interpretations rendered through formal agencyadjudications that were not self-executing.333

Understanding Chevron as a current preferences default rule can make sense of these apparentanomalies. While different forms of rulemaking have equally coercive effects on third parties, it is onlyrulemaking that is conducted after notice and comment that gives some reasonable assurance that theagency surveyed the current political preferences before acting. The notice-and-comment process alsoalerts Congressional members and the President’s political advisors that an issue is coming up that they maybe interested in influencing, or at least alerts the private parties who then alert these political officials.Likewise, formal adjudications require hearings that give the immediately affected parties an opportunityto be heard, and generally permit third parties to participate as well.334 A recent case apparently endeavorsto cut back on the prior doctrine requiring agencies to allow intervention by third parties.335 But in practiceformal adjudications provide affected parties with opportunities to make their political views known eitherdirectly or by alerting supportive members of the political branches. Thus, although some formaladjudications do not have the force of law in the formalistic sense, they do satisfy the conditions for a

336An alternative explanation is that when Congress authorizes agencies formally to adjudicate cases, it isexpressly delegating to the agency the authority to make those statutory interpretations that are incident to thatadjudication. Under this explanation, even if such agency adjudications would not by themselves fit within a currentpreferences default, Congress has opted out of that default rule by enacting the statutory provisions expressly providingfor such adjudication. The Court cases finding Chevron applicable to agency adjudications have in fact involved strongevidence that Congress was delegating interpretative authority to the agency adjudications. See, e.g., INS v. Aguirre-Aguirre, 526 U.S. 415, 424-25 (1999).

337 Mead, 121 S. Ct. at 2168-69 (quoting 19 U.S.C. §1502(a) and 19 CFR §177.9(a)).338 Id. at 2168-69.339 Chevron, 467 U.S. at 842, 863-64; Mead, 121 S. Ct. at 2181-82 (Scalia, J., dissenting) (“As Chevron itself held,

the Environmental Protection Agency can interpret ‘stationary source’ to mean a single smokestack, can later replacethat interpretation with the "bubble concept" embracing an entire plant, and if that proves undesirable can return againto the original interpretation.”)

340 Smiley v. Citibank, 517 U.S. 735, 742 (1996). See also Rust v. Sullivan, 500 U.S. 173, 186-87 (1991)(emphasizing that agencies “must” retain the power to change their interpretations over time).

341 See Mead, 121 S. Ct. at 2169, 2174-75.

85

current preferences default rule, and merit Chevron deference.336

Ordinary formalistic notions of when agency action has the force of law are also hard to square withMead’s result, which denied deference to a customs ruling letter that imposed a duty before certain goodscould be imported, even though by statute and regulation that ruling was “binding” on all Customs Servicepersonnel.337 The letter ruling thus had the concrete coercive effect of restricting importers’ freedom toimport the goods without paying a duty. Indeed, the effect was equivalent to a court injunction, which noone would deny has the force of law under ordinary understandings. The fact that the agency reserved theright to change its rulings in the future, although noted by the Mead Court,338 does not reduce theirformalistic force of law until changed, any more than a trial court injunction could be said to lack the forceof law just because it might be modified in the future by appellate courts or the trial court itself. Moreover,Chevron itself stressed the importance of allowing agencies the flexibility to change their interpretationsover time, and thus mandated deference even though the agency had changed its interpretation in the pastand might do so again in the future.339 “[C]hange is not invalidating, since the whole point of Chevron isto leave the discretion provided by the ambiguities of a statute with the implementing agency.”340 Anagency interpretation thus clearly need not be regarded as binding on the agency in the future to meritChevron deference.

The denial of deference in Mead instead rested on an entirely different notion of what the “forceof law” meant. That notion requires an agency position articulated as a generally applicable rule uniformlyapplied by the agency to a defined class of persons, as well as the serious solicitation of input, as throughnotice and comment rulemaking, by the members of the affected class. In addition to the lack of anynotice-and-comment period, the Court stressed that the agency rendered 10-15,000 letter rulings a yearthrough 46 separate branch offices.341 Further, agency regulations provided that third parties would notreceive notice of any change in letter rulings, and should not rely on them or “assume that the principles of

342 Id. at 2169. See also id. at 2174 (noting that “that a letter's binding character as a ruling stops short of thirdparties; Customs has regarded a classification as conclusive only as between itself and the importer to whom it wasissued”).

343 Id. at 2171, 2174 (emphasis added).344 Id. at 2174.345 This also explains something else difficult to explain under formalistic force of law notions: why decisions

by Administrative Law Judges that are appealable to higher levels in the agency should not get Chevron deference eventhough they can, unless appealed, independently command behavior and impose penalties. Merrill and Hickman gamelyargue that such decisions do not have the force of law because they do not bind the agency in the future. See Merrill& Hickman, supra note , at 908. But this seems inconsistent with their earlier definition of when actions have the forceof law, id. at 881, as well as with ordinary notions of when judicial decisions have the force of law, not to mentionChevron’s general willingness to mandate deference for interpretations the agency does not consider binding on itselfin the future. See supra at __.

346 See Mead, 121 S. Ct. at 2170.

86

that ruling will be applied in connection with any [other] transaction.”342 Such agency decisions, althoughhaving an individualized coercive effect on the recipients of the letter rulings, could not be said to have theforce of a generally applicable rule of law for a class of parties. Instead, Mead indicated that to getChevron deference the agency must have and be exercising the authority “to make rules carrying the forceof law,” which the Court concluded must “bespeak the legislative type of activity that would naturally bindmore than the parties to the ruling.”343 The Court did not mean to require formal rulemaking, since itrecognized that principles articulated in agency adjudication could suffice. But the Court did mean toexclude individualized arbitrations that bound only the parties to it and failed to create any general ruleapplicable in other cases. This is entirely understandable from a current preferences perspective. Suchindividualized determinations cannot be said to reflect the current political preferences of any general setof interested parties. One-shot individualized interpretations are also unlikely to provoke any legislativeoversight. They thus do not suffice to justify judicial application of a current preferences default rule, whichonce adopted would create an interpretation that would be binding on the general class of affected persons.

Mead thus offers an interesting lesson about uniformity. The fact that an agency position createsnational uniformity is not, as noted above, a sufficient condition for Chevron deference. Mead itself notedas much when it said that even when an agency ruling was precedent for other agency decisions, as withinterpretive rules, that did not suffice to confer Chevron deference.344 But Mead indicates that agencyuniformity is probably a necessary condition. This makes sense if the Chevron-Mead doctrine reflects acurrent preferences default rule. Agency uniformity may not reflect prevailing political preferences, and thusshould not suffice for deference. But uniformity is a necessary condition for Chevron deference because,if the agency position is not uniform across the nation, it cannot be considered an accurate manifestationof national political preferences.345 Indeed, a ruling by one of 46 agency interpreters that is not binding inother cases has serious uniformity disadvantages compared to interpretations by an appellate system,especially one that directed appeals through a nationwide Court of International Trade to a single FederalCircuit.346

Justice Scalia disagreed in a vehement dissent, taking the position that an authoritative (and thus

347 Mead, 121 S. Ct. at 2187-88 & n.6 (Scalia, J., dissenting).348 See supra at __; Mead, 121 S. Ct. at 2177 n.19.349 See Bowen, 488 U.S. at 212 (courts cannot accept “post hoc rationalizations” for agency action). 350 Mead, 121 S. Ct. at 2172-73& n.13 (citing Nationsbank v. Variable Annuity Life Ins., 513 U.S. 251, 256-57

(1995).).351 513 U.S. at 254-55. 352 Id. at 256-57.353 Mead, 121 S. Ct. at 2185 (Scalia, J., dissenting) (citing Mead Corp. v. Tilley, 490 U.S. 714 (1989), and Pension

Benefit Guaranty Corporation v. LTV Corp., 496 U.S. 633 (1990)). Actually, Justice Scalia claimed that the latter caseinvolved only a notice of restoration, but in fact the court pointed to three separate opinion letters that preceded thenotice. Id. at 642-643. Moreover, the notice of restoration itself might be regarded as an agency adjudication withinMead.

354 Tilley, 490 U.S. at 722 (case where three agencies – the Pension Benefit Guaranty Corporation, the IRS, andthe Department of Labor – all agreed on the relevant statutory interpretation in various opinion letters and guidelines

87

uniform) agency position had been taken because the agency head signed the brief defending theinterpretation in court.347 But the Court has never accepted that position,348 and seeing Chevron as acurrent preferences default rule helps us understand why. Part of the explanation, as noted above, is thatpositions taken in litigation briefs are less likely to reflect a solicitation of political views. But Meadillustrates a further point. If the Court had deferred to the interpretation, then in the future any one of the46 branch office interpretations would have to be considered binding in court if the agency head laterproved willing to defend it. Under this alternative rule, an agency head that generally committed herself toa strategy of defending interpretations by her 46 branch offices would create a system of agencyinterpretations that were binding even though there was no reasonable assurance they reflected currentpolitical preferences. The Chevron-Mead doctrine instead seems to require that the agency provide theassurance that its interpretations reflect prevailing national preferences before it imposes thoseinterpretations on others.349

Understanding Chevron as a current preferences default rule also helps provide some content toits residual “other” category of decisions deemed by Mead to have the “force of law.” The only examplethe Court cited was a “deliberative conclusion” by the Comptroller of the Currency.350 This decision hadtwo noteworthy differences from Mead. First, it was a top level decision by the head of the agency, undera system whereby the Office of the Comptroller of the Currency itself decided whether bank applicationsto expand their activities were “legally permissible.”351 Second, the agency head’s decision reflected his“deliberative conclusion” about what banks could legally do, formally reflected in his decision to grant thebank application, and was not taken during litigation.352 There was thus not the same concerns as wereraised in Mead about disuniform agency interpretations that were not based on any reflection of nationalpolitical preferences. Similarly, two other cases that Justice Scalia in dissent bemoaned were inconsistentwith Mead did involve opinion letters or guidelines.353 But they were not one-shot opinion letters that mightreflect agency disuniformity or a failure to test the interpretation against prevailing political preferences.Instead, they were cases involving multiple opinion letters and guidelines that clearly reflected a uniformnational position by politically accountable agencies, and were thus amenable to legislative oversight.354

written in advance of litigation); LTV, 496 U.S. at 642-643, 647-648 (case involving a consistent uniform agencyinterpretation evidenced by three separate opinion letters before the notice of restoration challenged in the case).

355 Mead, 121 S. Ct. at 2184-85 (Scalia, J., dissenting) (citing FDIC v. Philadelphia Gear Corp., 476 U.S. 426 (1986),and Young v. Community Nutrition Institute, 476 U.S. 974 (1986)).

356 See Philadelphia Gear, 476 U.S. at 436-39 (deferring to FDIC interpretation of a statutory “deposit” that hadbeen uniformly applied for over sixty years by the agency in interpreting its own regulations, and that Congress(knowing this was the prevailing interpretation) had reenacted and otherwise amended the statutory definition of adeposit, and even incorporated FDIC regulatory definitions); Young, 476 U.S. at 983 (longstanding uniformly appliedagency interpretation that Congress had not overturned even though it otherwise amended the relevant provision).

357 See supra Section III.A.358 Mead, 121 S. Ct. at 2172, 2174.359 See supra Section IV.B.360 Mead, 121 S. Ct. at 2172, 2174 & n.16. The court tried to get around this by noting that other statutory

provisions provided the reviewing court could consider new grounds and facts, id. at 2174 n.16, but this did not offerany response to Justice Scalia’s cogent critique that this statute provided no reason to deny deference on legal issuesthat had been considered by the agency, id. at 2187 (Scalia, J., dissenting).

361 Mead, 121 S. Ct. at 2181 (Scalia, J., dissenting).362 See Strauss, The Rulemaking Continuum, 41 DUKE L.J. 1463, 1475-87 (1992); Merrill & Hickman, supra note

, at 903 (collecting cases). Another possible definition is that interpretive rules interpret existing norms whereas

88

The only two other cases Scalia cited as inconsistent with Mead also involved longstanding uniform agencypositions.355 Moreover, they fell within the doctrine that courts should follow prevailing interpretations thatCongress was aware of failed to alter even though it re-enacted or amended the statute.356 As explainedabove, even if such an interpretation is not otherwise binding (here under Chevron), courts are still justifiedin adopting such an interpretation as the current preferences default rule.357

One could try to rationalize all these distinctions by referring to an implicit congressional intent. TheMead Court itself referred to claims about what Congress “would expect” or “meant” when it createdagency power that did not satisfy the Court’s “force of law” test.358 But efforts to explain Chevron byreference to enacting legislative intent are generally strained,359 no more so than in Mead itself where themost direct indicator of legislative intent was a statutory provision that cut the other way by providing thatthe agency decisions must be “presumed to be correct” on judicial review.360 Moreover, as Justice Scaliapointed out, if Chevron were truly based on an implicit congressional intent to delegate authority to theagency by enacting the relevant statute, then it would seem that the mere conferral of rulemaking authorityto resolve statutory ambiguities should suffice, and deference should be granted even if the agency choseto resolve statutory ambiguities through some other method.361 Understanding Chevron as a currentpreferences default rule provides a better explanation. By what Congress meant or expected the Courtinstead means what general default rule Congress would want, not what the enacting Congress intendedfor this statute in particular. And the reason the doctrine depends not just on how much power the agencywas granted, but on how the agency exercises its power, is that only certain methods of exercise providethe reasonable assurance that the agency action reflects current governmental preferences.

Also consistent with this theory is the Court’s otherwise puzzling distinction between interpretiveand legislative rules. An interpretive rule binds agency personnel but not the regulated community.362

legislative rules create new norms, see id., but that definition can hardly help distinguish cases meriting Chevrondeference since all such cases involve rules that interpret existing statutory norms.

363 Mead, 121 S. Ct. at 2174. See also Christensen, 529 U.S. at 587 (interpretive rules not entitled to Chevrondeference).

364 5 U.S.C. § 301 (2000). See generally Chrysler Corp. v. Brown, 441 U.S. 281, 308-12 (1979) (describing historyof "housekeeping statute").

365 5 U.S.C. § 558(b) (2000); ICC v. Cincinatti, New Orleans, & Tex. Pac. Ry. Co., 167 U.S. 479, 494 (1897);Chrysler Corp. v. Brown, 441 U.S. 281, 302-08 (1979); Merrill & Hickman, supra note , at 883.

366 See 5 U.S.C. §553; Merrill & Hickman, supra note , at 903. Interestingly, there appears to be a chicken-and-egg phenomenon at work. The reason the 1946 Administrative Procedure Act exempted "interpretative rules" fromnotice- and-comment requirements was that courts at the time did not defer to them. Id. at 887. The important point isthat the two go together: some opportunity for political participation (like notice-and-comment procedures) and deferenceto the result as likely tracking current political preferences.

367 See MICHAEL I. SALTZMAN, IRS PRACTICE AND PROCEDURE ¶ 3.02[3] (2d ed. 1991).368 Mead, 121 S. Ct. at 2174. Procedural rules are also exempt from notice-and-comment, see 5 U.S.C. §553(b)(A)

(2000), and thus should also not receive Chevron deference as a class. 369 See United States v. Haggar Apparel, 526 U.S. 380 (1999).

89

Unlike legislative rules, “interpretive rules . . . enjoy no Chevron status as a class.”363 As a matter oflegislative intent, this distinction makes little sense. If the enacting Congress intended to delegate anyauthority to agencies, it would be authority over the agencies’ own personnel. Less deference, if any,seems merited when the agency tries to exert authority over third parties. To the extent one can read anyintent from Congressional statutes, they also seem to cut against the prevailing distinction. The“housekeeping statute” of 1789 provides that: "[t]he head of an Executive department or militarydepartment may prescribe regulations for the government of his department, [and] the conduct of itsemployees . . .”364 This is an express and very general authority for agencies to adopt interpretive rulesbinding agency personnel. In contrast, both the Administrative Procedure Act and caselaw deny agenciesany general authority to issue legislative rules that bind third parties unless such authority has been expresslyprovided in other specific statutes.365 Thus, any implied legislative intent would support a doctrine of super-Chevron deference for interpretive rules rather than the actual doctrine of less deference. Under a currentpreferences default rule, in contrast, the distinction does make sense because interpretive rules do notrequire any notice-and-comment whereas legislative rules do.366 Thus legislative rules are promulgated afterthe sort of political participation that provides some assurance they track current political preferences, butinterpretive rules need not be. On the other hand, some agencies (like the Treasury Department) make ita practice to use notice-and-comment procedures when adopting interpretive regulations.367 If notice-and-comment is provided, such interpretive regulations do fit within a current preferences default rule and meritChevron deference, which probably explains the tradition of great deference to IRS interpretations andwhy the Mead Court did not say interpretive rules never get Chevron deference but only that they do notautomatically get it “as a class.”368

A current preferences default rule also explains better why Chevron deference is granted evenwhen a statute makes agency decisions subject to a trial de novo in court.369 Under the legislative intentmodel, this result has seemed odd because the enacting Congress signaled, if anything, a desire for non-

370 See Merrill & Hickman, supra, at 840-41, 871-72, 909. Merrill and Hickman also argue that any presumptionbased on presumed congressional intent should be weak to avoid distorting legislative intent, id. at 888, which is alsoinconsistent with the doctrine since in their view Haggar reflects a strong presumption.

371 Chevron, 467 U.S. at 845, 864-66.372 Id. at 845, 864-66.373 Bob Jones University v. United States, 461 U.S. 574 (1983); see discussion supra Section III.B. Although

Bob Jones was decided a year before Chevron, there were already tax cases providing for a similar rule of deference, andin any event the case illustrates the point.

374 See supra Section III.B.375 The Treasury interpretation was a change from an IRS interpretation that had prevailed since 1970, but such

changes in interpretation are perfectly proper under Chevron. See supra Sections IV.A-B. Interestingly, the Court didsuggest some willingness to defer to the IRS interpretation, which it stressed was consistent with current political views,see Section III.B, even though it seemed blind to the contrary interpretation by the IRS’s superiors in Treasury. Underthe theory urged here, this is explainable because the latter’s admitted lack of conformity to current political views gavethem no resonance.

90

deferential judicial review. The only explanation within this model has been that the presumption that thelegislature intends courts to defer to agencies is extremely difficult to rebut, which is hard to justify since thebasis for that presumption is acknowledged to be weak and even fictional.370 But as a current preferencesdefault rule, the result follows naturally. De novo review of the facts involves the exercise of a traditionaladjudicative function that does not require accurately ascertaining prevailing political preferences. It is thusnot at all in tension with the court deferring on how the agency fills statutory gaps, a result justified undera current preferences default by the agency’s advantage in ascertaining current enactable preferences.

Finally, a sometimes underappreciated limit is that Chevron deference only applies to agencyinterpretations that reflect a “reasonable policy choice.”371 The Court took pains to point out did not meana choice reasonable according to judge’s personal policy preferences but rather a reasonableaccommodation of competing political interests.372 This does mean, however, that to merit Chevrondeference the agency decision must be based on its policy or political judgment, not the agency’sdisagreement with the courts about hermeneutics or the proper default rule. The courts defer to agenciesbecause their policy views are more in tune with prevailing political preferences, not because agencies arebetter at legal interpretation. If the agency by its own account is not reflecting current political preferences,then its decisions do not merit deference under a current preferences default rule.

Bob Jones is illustrative.373 Although the then-current Treasury Department interpreted the statuteto allow tax exemptions for racially discriminatory schools, it made no claim its interpretation reflectedsound policy or current political preferences on whether racially discriminatory schools should get taxexemptions. To the contrary, the agency supported a new statute to expressly deny such tax exemptions.374

Instead, the Treasury interpretation was based on its view about statutory meaning, and its premise thatenacting rather than current political preferences should resolve any ambiguity. The Court gave noindication it even considered deferring to this interpretation.375 If Chevron were based on an implicitenacting legislative intent to delegate to the agency the authority to resolve ambiguities, this is hard tounderstand since, if such a delegated power existed, the Treasury Department was exercising it.

376 See ESKRIDGE, DYNAMIC, supra note, at 164-70; Farina, supra note , at 501-26; William N. Eskridge, Jr. & JohnFerejohn, The Article I, Section 7 Game, 80 GEO. L. J. 523, 528-40, 547-51 (1992); William N. Eskridge & John Ferejohn,Making the Deal Stick , 8 J. L. ECON. & ORG. 165 (1992). Others share the premise but celebrate it rather than critique it.See Mashaw, Prodelegation: Why Administrators Should Make Political Decisions, 5 J.L. ECON. & ORG. 81, 98 (1985).

91

Understanding Chevron instead as a current preferences default rule, in contrast, readily explains why theSupreme Court would ignore a Treasury position that made no pretense of reflecting current politicalpreferences. The courts are not looking to agencies for expertise in how best to conduct legal interpretationof administrative statutes, but rather for their expertise in where current political preferences lie to resolveambiguities left by legal interpretation.

Thus, courts should not, and do not, give Chevron deference to agency interpretations thatadmittedly do not match current enactable policy preferences. This requirement that the agency make a“reasonable policy choice” given competing political interests also suggests courts should deny deferenceif, even though not openly admitted by the agency, its interpretations plainly conflict with current enactablepolicy preferences. This is obviously an exception that must be construed narrowly lest it swallow the rule.If courts in every case make their own assessment of where current enactable preferences lie, and rejectevery agency interpretation that conflicts with that assessment, there would be no effective deference. Norwould that advance political satisfaction because the entire premise of Chevron is that agencies generallyknow better than courts where current enactable preferences lie. However, there is a definable set ofextraordinary cases where a court can determine with confidence that the agency interpretation conflictswith current enactable preferences. But before reaching those cases, we need to confront an issue that hasso far been deferred: what should courts do when the political views of the executive and legislatureconflict?

D. Conflicts Between the Executive and LegislatureThe affirmative argument above for Chevron deference depends on agencies being a fairly accurate

barometer of current enactable preferences, or at least more accurate than judicial estimates would be.Although the phrase “legislative preferences” is often used as a shorthand, in systems like the United Statesthe set of preferences that are actually enactable is strongly influenced not just by legislators but also by theexecutive who has powers to veto and put issues on the public agenda, each of which is in turn influencedby outside political forces who lobby, organize, contribute and otherwise constrain the actions of bothlegislators and executives. In such a system, agencies must be responsive to executive, legislative, andgeneral political influences for their actions to be likely to minimize the dissatisfaction of enactable politicalpreferences in a way that merits Chevron deference.

Academic critique of Chevron has tended to presume that agencies respond to Presidentialpreferences but not congressional preferences, and that Chevron thus alters the constitutional balance ofpower by shifting power from Congress to the President where their preferences conflict.376 Where thepremises of this argument hold, this is a powerful critique. But the premises are not generally applicablefor two reasons.

First, most agency decisions that get Chevron deference do not involve dramatic conflicts

377 Cf. Elhauge, Preference-Eliciting, supra note , at X.A. (noting that various pro-federalism canons can alsobe justified on the related theory that, when enactable national preferences are unclear, the rule of thumb most likely tominimize political dissatisfaction is to track the preferences of political subunits like states).

378 See supra Section IV.A. Because Eskridge and Ferejohn assume agencies track Presidential preferences, theyare also led to the conclusion that the legislative veto and delegation of administrative powers to legislative agencieswere both proper responses to the administrative state that merely restored the original political balance of powerbetween the President and Congress, and that thus both INS v. Chada, 462 U.S. 919 (1983), and Bowsher v. Synar, 478U.S. 714 (1986), were wrongly decided. Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 540-47. If, as argued here,agencies in fact are responsive to both the executive and legislative branches, then legislative veto statutes and thedelegation of power to legislative agencies both try to alter the balance by making agencies responsive only to thelegislative branch, thus effectively circumventing the President’s constitutional veto power. The analysis here can thusexplain Chada and Bowsher as well as Chevron. Other rational choice literature also generally incorporates anassumption of agency-Presidency correspondence, and thus not surprisingly produces different conclusions than myanalysis here. See infra Section VII.E.

379 See supra Section IV.A; Chevron, 467 U.S. at 866 (“agency . . . may . . . properly rely upon the incumbentadministration's views . . . . While agencies are not directly accountable to the people, the Chief Executive is . . . “)

380 Kagan, Presidential Administration, 114 HARV. L. REV. 2245, 2250, 2290-94, 2323 & n. 308 (2001); RichardH. Pildes & Cass R. Sunstein, Reinventing the Regulatory State, 62 U. CHI. L. REV. 1, 25 (1995); Gary Lawson, The Riseand Rise of the Administrative State, 107 Harv. L. Rev. 1231, 1242-46 (1994); Thomas O. Sargentich, The Administrative

92

between the executive and legislative branches. The real world is notably less exciting. Most agencydecisions involve issues too minor to interest the legislature or for which there is no clear indication wherelegislative preferences lie. This does not mean such cases are unimportant. The entire point of preference-estimating statutory default rules is to resolve the interstitial gaps or ambiguities where legislative action isunlikely. Indeed, it is precisely in such cases that the default rule matters most because that is when thedefault sticks and becomes enduring law. In this typical set of cases where there is no other clear indicatorof legislative preferences, an agency decision responsive to Presidential preferences may be the bestavailable indicator of where enactable preferences would lie if the issue ever came to a vote. ThePresident is, after all, responsive to the same electorate as Congress as a whole, and while the differentelectoral rules for registering that electorate’s preferences in Presidential and congressional elections willlead to many differences, it should also lead to broad areas of likely agreement. Absent any other indicatorof legislative preferences, then, tracking Presidential preferences seems likely to minimize politicaldissatisfaction.377

Second, and more important, the premise of agency-executive equivalence that underlies thiscritique is contradicted by an extensive literature showing that agencies in fact are subject not only toPresidential influence but to various forms of influence by Congress and outside political groups.378 To besure, the academic critique finds support in the language of Chevron itself, which seemed to base thedeference rule on agencies’ accountability to the current President rather than to the political branches asa whole.379 But if so understood, these statements reflected a very superficial understanding about therelationship between agencies and the President. Not only did the above forms of de facto influence fromCongress and outside political interests clearly exist, but at the time the President was understood to lacklegal authority to direct how agencies should exercise discretionary authority delegated to those agenciesby legislation.380 Any implication to the contrary in Chevron was just dicta about a topic that the Court

Process in Crisis-- The Example of Presidential Oversight of Agency Rulemaking, 6 ADMIN. L.J. 710, 716 (1993); PeterL. Strauss, The Place of Agencies in Government: Separation of Powers and the Fourth Branch, 84 COLUM. L. REV. 573,649-50 (1984) .

381 See Chevron, 467 U.S. at 853-59, 864. The Court also stressed that enacting legislative preferences wereunclear. See id. at 845 (“Congress did not actually have an intent” on the relevant question). The less clear enactingpreferences are, the less reliable the evidence of correspondence with current preferences has to be to merit deference.See supra II, III.B.

382 See Kagan, supra note , at 2347 (noting empirical evidence that independent agencies are more responsiveto Congress than executive agencies are).

383 Id. at 2376-77 (proposing such a distinction but acknowledging it does not reflect current law). Althoughresponsive to congressional preferences, independent agencies are also responsive to Presidential political preferences.See Moe, Regulatory Performance and Presidential Administration, 26 AM. J. POL. SCI. 197, 220-21 (1982).

384Cf. Kagan, supra note , at 2373-75 (acknowledging that various limits on Chevron would not make muchsense if the doctrine were based on accountability to the President).

385 See supra Section IV.C.386 See Kagan, supra note , at 2281-2319.387 Id. at 2312.388 I take no position here on whether such Presidential authority over agencies is constitutionally protected

or prohibited, and normatively desirable or not. See id. at 2251, 2319-63 (discussing these issues). My only point here

93

did not directly consider since the case did not involve any clear clash between the President and Congress.Although the new agency interpretation considered in Chevron did come in response to a change ofadministration, the Court stressed that the interpretation that the agency was abandoning had not reflectedany prior understanding about current enactable policy preferences, but was rather a mechanical responseto a lower court decision’s legal interpretation.381 Indeed, if Chevron were based on accountability toPresidential views, then independent agencies should get less deference than executive agencies that aresubject to agency removal power.382 In fact, Chevron draws no such distinction.383

Moreover, the various doctrinal limits detailed above mean Chevron deference cannot in fact betriggered by mere agency responsiveness to executive preferences. If such responsiveness sufficed, thennone of the Chevron limits should apply because all agency decisions potentially have the requisiteaccountability to the executive.384 Instead, the doctrines that limit Chevron require broader indicia that theagency was accurately informed about current political preferences outside the executive branch beforeit took action.385 The analysis here suggests a further limit on Chevron deference: the agency action cannotjust reflect executive views that plainly conflict with the legislature’s. With these limits in place, Chevronmakes sense as a general current preferences default rule.

Different administrations might press against those limits in different ways. My colleague and formerClinton policy advisor Elena Kagan has shown in her recent masterpiece on Presidential administration thatthe Clinton Administration aggressively expanded Presidential authority over agencies, includingappropriating agency action and assuming the power to direct it.386 Further, “Clinton's use of directiveauthority over the executive branch agencies accelerated dramatically when the Democrats lost control ofCongress . . .”387 Such a shift would, if my current preferences default rule theory is right, naturally provokea corresponding shift in the application of Chevron deference.388 If agency action can largely be directed

is that, if such Presidential authority is exercised in ways that mean agency action does not reflect current enactablepreferences, then judicial deference based on likely agency correspondence to those preferences should be denied.Professor Kagan takes the contrary view that Presidential involvement should actually heighten, rather than weaken,Chevron deference. See id. at 2372-80. But this is based on her adopting the premise I rejected in Section I: that wherecongressional intent is unclear, the default rule should reflect what she (or judicial judgment) regards as the wisestpolicy, id. at 2330-63, even though that conflicts with enactable preferences, id. at 2345 (acknowledging that “bothPresidents [pursuing her approach] self-consciously put in place a set of policies that could not have succeeded ingetting through Congress”). Professor Kagan also acknowledges that her position does not fit the actual contours ofthe Chevron doctrine. Id. at 2373-76.

389 See id. at 2283.390 Id.391 Id.392 120 S. Ct. 1291 (2000). 393 Id. at 1303-04, 1306-13.394 Id. at 1307-08, 1310-11.

94

by the President, then its actions are more likely to reflect only Presidential preferences. Such agencyaction is more likely to take positions that create sharp conflicts with Congress. More important, suchagency action is less likely to reflect current enactable preferences since the President’s views reflect onlyone piece of the political puzzle necessary to create enactments. Agency action that just takes one side ina heated political disagreement between the executive and legislative branch should under my theory thusbe denied Chevron deference.

Which is precisely what we have seen. Two recent cases dramatize the shift in Chevron that hastaken place in reaction to the Clinton expansion of presidential authority over agencies. One of theprominent examples of Presidentially-led agency action given by Professor Kagan involves PresidentClinton’s announcement that as executive he would restrict the marketing of cigarettes through FDAregulations. 389 This announcement came before any FDA rule had actually been adopted and indeedbefore any notice and comment period had begun, but nonetheless stated the regulatory principles that thePresident said “will” be adopted “by executive authority,” and the FDA later adopted precisely thoseprinciples.390 Further, the President announced that he was “authorizing” the FDA regulations,391 thuspresupposing an answer to the contested question of statutory interpretation over whether the FDA’sstatutory authority to regulate drugs extended to cigarettes.

This statutory issue ultimately was resolved in FDA v. Brown & Williamson,392 where theSupreme Court rejected the FDA’s interpretation that cigarettes were a “drug” or “device” within themeaning the Food, Drug, and Cosmetic Act. The Court did not claim that this agency interpretationconflicted with the plain meaning of the statute. Instead, the Court relied mainly on a series of subsequentstatutes that retained or relied on the FDA’s lack of authority to regulate tobacco, opting instead for a morelimited regulatory strategy of requiring warnings, limiting advertising, and banning sales to minors.393 TheCourt also noted that the agency itself had just recently altered its longstanding interpretation that the Actdid not give it authority to regulate cigarettes.394 But the Court took pains to stress that its “conclusion doesnot rely on the fact that the FDA's assertion of jurisdiction represents a sharp break with its prior

395 Id. at 1313. See also id. at 1328-30 (Breyer, J, dissenting, joined by Stevens, Souter, and Ginsburg, JJ.)(collecting cases and quoting Chief Justice Rehnquist for the proposition that: “A change in administration broughtabout by the people casting their votes is a perfectly reasonable basis for an executive agency's reappraisal of the costsand benefits of its programs and regulations.”) The Court’s ambivalent reliance on the change in agency interpretationmay also have reflected a failure to distinguish between Chevron and Skidmore deference, see supra Section IV.B, orperhaps even between hermeneutic inquiry and default rule analysis. Where meaning is unclear and must be filled witha default rule, agency change is fine and indeed encouraged. But this is not inconsistent with the hermeneutic point thatthe agency’s contemporaneous or past interpretation might be deemed relevant to establishing a statutory meaning fromwhich the current agency was not free to deviate. On this point, the Court could have cited Bankamerica Corp. v. UnitedStates, 462 U.S. 122, 131 (1983) (“the mere failure of administrative agencies to act is in no sense ‘a binding administrativeinterpretation’ that the Government lacks the authority to act. However, ‘just as established practice may shed light onthe extent of power conveyed by general statutory language, so the want of assertion of power by those whopresumably would be alert to exercise it, is equally significant in determining whether such power was actuallyconferred.’”) (collecting cases); General Electric v. Gilbert, 429 U.S. 125, 140-45 (1976) (declining to follow an agencyguideline on statutory meaning when it was not contemporaneous with enactment, and it changed many years later)(collecting other cases with similar rulings). Whether Chevron overrules this line of cases as a hermeneutic matter aswell as a default rule matter is an interesting question. My answer would be ‘no’ because, if the statute has a fixedmeaning, then varying agency interpretations indicate the agency has sometimes wrongly divined it, and that theinterpretation closest to the creation of meaning is most likely to be valid. The author of Chevron himself seems to agree.See Federal Election Commission v. NRA Political Victory Fund, 513 U.S. 88, 103 (1994) (Stevens, J, dissenting).

396 FDA, 120 S. Ct. at 1313.397 Id. at 1327 (Breyer, J., dissenting) (observing that “Congress both failed to grant express authority to the

FDA when the FDA denied it had jurisdiction over tobacco and failed to take that authority expressly away when theagency later asserted jurisdiction.”)

398 Id. at 1329-30 (Breyer, J., dissenting).

95

interpretation,” recognizing that this would conflict with Chevron.395 Rather, the important point to theCourt was that several intervening Congresses had effectively incorporated the agency’s prior interpretationinto a broad political compromise leading to several subsequent statutes.396

The Court’s objection, then, was not so much that the FDA was changing its interpretation, but thatin doing so it was deviating from the most recent official indications of enactable political preferences. TheCourt was well aware that the FDA’s changed interpretation was vigorously championed by PresidentClinton and vociferously opposed by then-current Republican majority in Congress. It was also aware that,given this deadlock, “Congress could not muster the votes necessary either to grant or deny the FDA therelevant authority.”397 Sustaining the FDA interpretation would thus amount to allowing the agency (andPresidency) to dictate the resolution of a political fight between two factions, neither of which weresufficiently powerful to procure legislation on their own. The Court rejected Justice Breyer’s dissentingposition that political accountability to the President should suffice when such a political deadlock exists.398

Instead, the Court in effect relied on the most recent indications it had about what could sufficiently satisfypolitical preferences to procure legislation: the preferences of intervening congresses reflected in subsequentlegislation. Had the current Congress instead supported the FDA’s changed interpretation, the case wouldlikely have come out the other way.

The Court explained its non-application of Chevron under what has become known as the“extraordinary case” exception.

399 FDA, 120 S. Ct. at 1314.400 See Merrill & Hickman, supra, at 844-45, 912-13.401 Mississippi Power & Light v. Mississippi ex rel. Moore, 487 U.S. 354, 381 (1988) (Scalia, J., concurring in

judgment); see Ry. Labor Executives' Ass'n v. Nat'l Mediation Bd., 29 F.3d 655, 676-77 (D.C. Cir. 1994) (en banc) (Williams,J., dissenting); Quincy M. Crawford, Comment, Chevron Deference to Agency Interpretations That Delimit the Scopeof the Agency's Jurisdiction, 61 U. CHI. L. REV. 957 (1994).

402 The concern motivating a scope of jurisdiction exception, although really applicable to all agencyinterpretations, is a real one: that agencies are biased in favor of expanding their own power. See supra Section IV.B.But the mechanisms for keeping this bias in check are the various forms of political accountability that make agenciesrelatively reliable determinants of current political preferences.

403 FDA, 120 S. Ct. at 1315-16.404 119 S. Ct. 765 (1999).405 Id. at 775-78.406 Id. at 777 (admitting linguistic ambiguity but finding it resolved by historical context); id. at 780-81 (statutory

intent at least not clear).407 Id. at 778.

96

Deference under Chevron to an agency's construction of a statute that it administers is premisedon the theory that a statute's ambiguity constitutes an implicit delegation from Congress to theagency to fill in the statutory gaps. In extraordinary cases, however, there may be reason tohesitate before concluding that Congress has intended such an implicit delegation.”399

Some have viewed this as a possible adoption of the much-debated claim that courts should not defer toagency interpretations about the scope of their own jurisdiction.400 But the problem with that claim is thatevery statutory interpretation implicates the scope of agency jurisdiction by defining what comes within thestatutes over which the agency has undenied jurisdiction.401 A “scope of jurisdiction” exception thus doesnot helpfully distinguish a set of cases that differ from others where deference is warranted.402 But whatthe Court mainly pointed to as showing that agency interpretation was “extraordinary” was that: (1) it wasan issue important enough for Congress to have preferences and (2) the agency action conflicted with thepolitical views reflected in a series of Congressional actions subsequent to the initial enactment.403 That is,the case involved precisely the two premises that the analysis above suggested would undermine Chevrondeference. Thus, one way to provide more concrete content to the vague “extraordinary case” exceptionwould be conclude that cases are “extraordinary” if they involve a split on a politically salient issue betweenthe President and Congress that makes it clear that the agency position does not reflect any enactablepolitical preference.

In Department of Commerce v. United States House of Representatives,404 the question waswhether the census statute could be interpreted to allow statistical sampling in apportioning congressionalseats. The statute, in various amended versions, had been interpreted to prohibit statistical sampling from1790 until 1994, when the Census Bureau changed its interpretation.405 The majority agreed the matterwas not settled by the plain language of the statute.406 Nonetheless, the agency waived any claim toChevron deference in light of its changed position.407 This might seem odd since Chevron expresslysanctions changes in agency interpretation. But the agency no doubt recognized its tenuous political

408 Id. at 787 (Stevens, J., joined by Souter & Ginsburg, JJ., dissenting) (acknowledging that the use of samplehad become a “partisan issue”).

409 Id. at 775-77. Three justices also relied on a general assessment that legislators would not prefer to changethe method by which they are elected without some discussion. Id. at 779 (O’Connor, J., Plurality Opinion as to III.B,joined by Rehnquist, C.J. & Kennedy, J.).

410 See supra Part II.411 As always, one should distinguish the default rule case (where the statute is ambiguous) from the case where

there is a clear statutory meaning to delegate authority to the President or agency. In the latter case, one does not reachthe default rule issue, and (absent some constitutional obstacle) courts are bound by the President or agency’s exerciseof statutory authority even if it conflicts with current congressional preferences.

412 Of course, the unenacted Congressional bill would not be entitled to deference either. Instead, since nocurrent preference could be said to be enactable regarding the interpretive choice, the court must turn to the most recentindication of enactable preferences, or perhaps to a preference-eliciting default rule.

413 See Kagan, supra note , at 2378 (if Chevron is based on Congress’ intent to delegate, then Presidentialinvolvement should make no difference).

97

position given that the change reflected its decision to side with President Clinton on an issue that wasstrongly opposed by the then-current Republican Majority in Congress.408 Thus, any choice of default rulecould not be overturned by the political process: the Republican Congress would block any statute thatauthorized statistical sampling; and the Democratic President, coupled with a veto-sustaining congressionalminority, would block any statute that took such authority away. Deferring to the agency on the default rulewould thus give it the power to take sides in this political stalemate. The agency did not claim such power,perhaps understanding the implicit limits on Chevron, and the Court instead resolved the matter with itsassessment of the legislative preferences that actually sufficed to enact prior versions of the statute.409

This is consistent with the point established above that, when there is no reliable basis fordetermining current or recent enactable preferences, the proper tack is to rely on the best estimate ofenacting legislative preferences.410 That will generally be the case when the current political branches areat loggerheads on some issue. Tracking current enactable preferences is not the same as deferring toPresidential Administrations; nor is it an invitation to pretend a sufficient political consensus exists when itplainly does not.411 A court can often do no better than make an educated estimate about what is currentlyenactable, or adopt a general rule of deferring to agencies who can make those estimates better than courts.But when a court knows that a particular interpretation could not get enacted in the current legislature, thenit should not be adopted as the interpretation that best estimates current enactable preferences. This wastrue in the above cases. It would also be true if Congress enacted a bill to overrule an agency interpretationbut was vetoed by the President. An agency interpretation protected from legislative overrule only byexecutive veto should not get Chevron deference.412 Under Chevron the courts adopt a general rule ofdeferring to an agency that has been duly influenced by Presidential, legislative and outside political input.The courts do not defer to the President himself, who reflects only one piece of the political puzzlenecessary to create an enactable political preference. Again, this is a hard limit to explain if Chevron werereally based on the conclusion that the enacting Congress intended to delegate interpretive authority, sincethe agency would be exercising delegated authority either way.413 But it does make sense if Chevron is

414 Since this merely states a default rule, these same limits would not apply if Congress expressly delegatedinterpretive authority to the agency.

415 If legislative reaction seems likely, then highly uncertain enactable preferences may be grounds for insteadusing a preference-eliciting default rule to better ascertain enactable preferences, but that topic is discussed in Elhauge,Preference-Eliciting, supra .

416 ESKRIDGE, DYNAMIC, supra note, at 167-70; Linda Cohen & Matthew Spitzer , Solving the Chevron Puzzle,57 LAW & CONTEMP. PROBS. 65, 73-76 (Spring 1994); Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 528-40, 547-51;Eskridge & Ferejohn, supra note , 8 J. L. ECON. & ORG. at ___.

417 See supra Section II.D (collecting sources).418 Id.419 ESKRIDGE, DYNAMIC, supra note, at 164-70; Cohen & Spitzer, supra note , at 69; Eskridge & Ferejohn, supra

note , 8 J. L. ECON. & ORG. at ___; Eskridge & Ferejohn, supra note , 80 GEO. L.J. at 533, 535-40. The text in these articles

98

a current preferences default rule.

E. Conclusion on Agency Deference.In short, Chevron does not require deference to all agency action, but limits that deference in many

ways.414 The courts defer on agency gap-filling, not on hermeneutic interpretation of whether statutorymeaning leaves a gap. The courts defer on agency gap-filling arrived at after general political input, not onagency positions insulated from such outside input. And the courts refuse to follow agency interpretationsthat either contravene prevailing political preferences or fashion a false political consensus out of what isin fact open political gridlock.

None of this means that the deference provided within the limits of Chevron rests on any unrealisticview that agencies perfectly reflect enactable preferences. Enactable preferences may be highly uncertainand agencies may often fail to reflect them. To justify Chevron as a current preferences default rule, it needonly be the case that, within Chevron’s limits, agencies are more likely to reflect enactable preferences thanjudicial judgment would. If judicial judgment is 10% likely to reflect an enactable preference, and agencydecisions are 20% likely to reflect an enactable preference, then a general policy of following agencydecisions will still maximize the satisfaction of enactable preferences.415 But where judges have specialreason to know the agency action has 0% odds of reflecting an enactable preference, the court should notdefer. More generally, the nuanced doctrines that set the limits on Chevron can properly be understoodas excluding those cases where there is little reason to think the agency action has any comparativeadvantage in reflecting enactable preferences.

These conclusions about Chevron differ greatly from those offered in some terrific prior work byProfessors Eskridge, Ferejohn, Cohen and Spitzer.416 As foreshadowed above, this difference results froma difference in assumptions. Unlike my analysis, these models assume judges just try to maximize their ownpreferences.417 But the difference in assumptions more specific to the Chevron issue is that these modelsassume that executive and legislative preferences are perfectly known.418 This assumption is plainlyincorrect, and ignores the possibility that agencies can help courts ascertain current enactable preferences,which I show above is actually the central purpose of Chevron. These models also wrongly assumed thatagencies only reflect executive preferences,419 when in fact agencies are also strongly influenced by

sometimes acknowledge the legislature also exerts some influence on agencies but argue it is small, see, e.g., id. at 540,and then in the actual models the articles assume agency action reflects only Presidential preferences.

420 See supra IV.A, D.421 ESKRIDGE, DYNAMIC, supra note, at 165-70; Cohen & Spitzer, supra note , at 69-70, 72-74; Eskridge &

Ferejohn, supra note , 80 GEO. L.J. at 536-40; Eskridge & Ferejohn, supra note , 8 J. L. ECON. & ORG. at ___.422 See Elhauge, Preference-Eliciting, supra note .

99

legislative preferences.420 Finally, these models wrongly assumed, contrary to the cases noted above, thatChevron deference applies when the agency position sides with the executive in a conflict with plainlegislative preferences.421 Indeed, because they ignore the informational feature of agency action, thesemodels assume that agency action only matters when it poses such a conflict. In contrast, under myanalysis, and current law, agency action only matters when it doesn’t pose that conflict and to the contraryhelps the court ascertain current enactable preferences.

VIII. CONCLUSION

Where a statute is ambiguous, and current enactable preferences cannot reliably be ascertained orelicited, an honest interpretive agent should endeavor to reconstruct the enacting government’s politicalpreferences to estimate of what it would have wanted to do. Sometimes this is relatively easy because anygovernment should share the preference to avoid absurd results or statutory invalidity, or because thestatute itself evidences the relevant preferences. But other times the preference estimate will be quiteuncertain. Nonetheless, the enacting government itself would want the court to proceed based on aprobabilistic estimate of its preferences. That estimate should also be based not only on the fragments theenacting government left in the legislative history of the particular statute, but on all sources of informationabout its preference structure at the time. Relying on enacting preferences does not mandate staticinterpretation, for those initial preferences may well call for changing interpretations as circumstanceschange. Further, where no one option seems more likely than not to reflect governmental preferences, thecourts should adopt the most moderate interpretive option. This helps explain many practices and seeminginconsistencies in statutory interpretation.

But a honest interpretive agent for the enacting government will also want to adopt a statutorydefault rule that tracks reliable official indications of the political preferences of the current governmentbecause the enacting government itself would want that general default rule applied to all statutes -- pastand future. This does not condemn judges to some crude model of political polling. A current preferencesdefault rule will only track reliable official indicia of enactable preferences, not transient political views thatare never shared by all the political actors necessary to enact statutes. Further, relatively categorical rules,like deferring to certain classes of agency interpretations, can be used to further this goal. Where greatuncertainty remains about enactable preferences, that itself will often call for using a preference-eliciting ruleinstead.422

To many the thesis of this article will seem highly counterintuitive: why wouldn’t the enacting

100

government want its own political preferences satisfied? The key is that the enacting government does, butcares not only about the interpretation of the statutes it enacts but also about the interpretation of olderstatutes being applied during its time of governance. To others, the analysis may seem sufficientlycompelling to suggest extending the proposition to updating defined statutory meanings. But allowingenacted statutory meanings to be updated by even relatively accurate estimates of current enactablepreferences circumvents a demanding constitutional process with a less demanding one, and reduces theaccuracy with which any estimate of enactable preferences can be made.

The analysis here differs sharply from typical rational choice models, which assume judges interpretstatutes to maximize the satisfaction of either (in most models) judicial preferences or (in some models) thepreferences of the enacting legislature regarding the particular statute being interpreted. These modelsignore the possibility that judges might, acting as honest agents for the enacting legislature, realize that theenacting legislature cares about the general default rule applied to all statutes, not just those it enacts. Alsocontrary to these models, I do not assume judges have anything resembling perfect knowledge of legislativepreferences, but rather assume they use certain methods of statutory construction precisely because theyhelp estimate those uncertain preferences.

My analysis also has the advantage of providing an explanation for many troubling twists and turnsin the statutory interpretation, especially within the Chevron doctrine. It also explains why there might belimits on the ability of legislatures to opt out of a current preferences default rule.

I began with the observation that our methods for resolving statutory meaning affect how oftendefault rule issues arise. But a feedback effect runs in the reverse direction too. The better our defaultrules, the less courts may feel impelled to insist on strained claims that hermeneutic methods establish aparticular meaning or intent in unanticipated cases. The judges most likely to feel the need to strain to findunambiguous meaning are those who -- quite properly -- believe that statutory interpretation shouldconstrain judges to effectuate legislature preferences. A theory of default rules designed to do preciselythat should increase their willingness to rely on default rules and thus lessen the cases resolved onimplausible claims that the statute has a determinate meaning.

In short, judges can be honest interpretive agents for the political process without always claimingthat the statutes that process has left us establish a clear meaning they can decode. To the contrary, thepolitical process should prefer to appoint interpretive agents who, rather than insisting on resolutions ofambiguous statutory meaning that are arbitrary or heavily tinged by (perhaps unspoken) judicialpreferences, resolve those statutory ambiguities with default rules that are designed to maximize thesatisfaction of political preferences.

By the same token, the more constraining our default rules, the less those who celebrate judicialjudgment over political judgment may feel an impetus to strain to find ambiguities in statutory meaning. Forinstead of sanctioning the exercise of judicial judgment, the proposed default rules would constrain judicialjudgment to further the satisfaction of political preferences. A well-developed theory of such default rulescan thus help avoid strained hermeneutic efforts in both directions.