Utah Law Review 2012 Number 1

TABLE OF CONTENTS ARTICLES Law, Fact, and Discretion in the Federal Courts: An Empirical Study Robert Anderson IV 1 Bargained Justice: Plea-Bargaining's Innocence Problem and the Brady Safety-Valve Lucian E. Dervan; Recantations Reconsidered: A New Framework for Righting Adam Heder & Wrongful Convictions Michael Goldsmith; Remedying the Misuse of Nature Sanne H. Knudsen; Due Process Denied: The Forgotten Constitutional Limits on Choice of Law in the Enforcement of Employee Covenants Not to Compete David A. Linehan; The Penguin and the Cartel: Rethinking Antitrust and Innovation Policy for the Age of Commercial Open Source Stephen M. Maurer; Class Action Defendants' New Lochnerism Mark Moller; A Minimalist Approach to Same-Sex Divorce: Respecting States that Permit Same-Sex Marriages and States that Refuse to Recognize Them Robert E. Rains; The Common Law of Disability Discrimination Mark C. Weber; NOTES E-Verify During a Period of Economic Recovery and High Unemployment Emily Patten; Incentive for Innovation or Invitation to Inhumanity?: A Human Rights Analysis of Gene Patenting and the Case of Myriad Genetics Laurie E. Abbott

UTAH LAW REVIEW Volume 2012 Number 1 UTAH LAW REVIEW Vol. 2012 No. 1 Pgs. 1 To 525 ARTICLES Law, Fact, and Discretion in the Federal Courts: An Empirical Study Robert Anderson IV Bargained Justice: Plea-Bargaining's Innocence Problem and the Brady Safety-Valve Lucian E. Dervan Recantations Reconsidered: A New Framework Adam Heder & for Righting Wrongful Convictions Michael Goldsmith Remedying the Misuse of Nature Sanne H. Knudsen Due Process Denied: The Forgotten Constitutional Limits on Choice of Law in the Enforcement of Employee Covenants Not to Compete David A. Linehan The Penguin and the Cartel: Rethinking Antitrust and Innovation Policy for the Age of Commercial Open Source Stephen M. Maurer Class Action Defendants' New Lochnerism Mark Moller A Minimalist Approach to Same-Sex Divorce: Respecting States that Permit Same-Sex Marriages and States that Refuse to Recognize Them Robert E. Rains The Common Law of Disability Discrimination Mark C. Weber NOTES E-Verify During a Period of Economic Recovery and High Unemployment Emily Patten Incentive for Innovation or Invitation to Inhumanity?: A Human Rights Analysis of Gene Patenting and the Case of Myriad Genetics Laurie E. Abbott i UTAH LAW REVIEW VOLUME 2012 NUMBER 1 The Utah Law Review was founded in 1948 with the purpose of serving the interests of students, the bench, and the bars of the state and surrounding areas. Since then, its scope has expanded to include legal issues of importance both domestically and internationally. The Utah Law Review is a student-run organization, with all editorial and organizational decisions made by student-editors enrolled at the S.J. Quinney College of Law at the University of Utah. The Utah Law Review publishes four issues per year, with each issue containing approximately 250 pages of legal scholarship. The Utah Law Review makes every effort to conform to the following editing sources: THE BLUEBOOK: A UNIFORM SYSTEM OF CITATION (19th ed. 2010) and THE CHICAGO MANUAL OF STYLE (16th ed. 2010). The Utah Law Review Board of Editors strives to maintain the text of published articles in its original format by editing only for accuracy and, where necessary, clarity. Views expressed herein are to be attributed to their authors and not to the Utah Law Review, its editors, the S.J. Quinney College of Law, or the University of Utah. COPYRIGHT: © 2012 by the Utah Law Review Society. Printed Spring 2012 (ISSN 0042- 1448). In addition to the rights to copy that are granted by law, the owner of the copyright of each article published in this issue, except as expressly otherwise noted, grants permission for copies of that article to be made and used for educational purposes by nonprofit educational institutions, provided that the author and journal are identified, that proper notice of copyright is affixed to each copy of the article, and that copies are distributed at or below cost. ACKNOWLEDGEMENTS: The editors of the Utah Law Review wish to express their gratitude to the following: The University of Utah S.J. Quinney College of Law, the Journal Alumni Association and the University of Utah Publications Council for their support; Professor Amy J. Wildermuth, faculty advisor; Angela Turnbow, editorial assistant; and the S.J. Quinney College of Law faculty and staff. Printed in the United States of America by Joe Christensen, Inc. ISSN 0042-1448 ii UTAH LAW REVIEW VOLUME 2012 NUMBER 1 SUBMISSIONS: The Utah Law Review receives submissions in March and April. Submissions should include a cover letter, providing the title of the work and the author's contact information (phone number and e-mail address). Submissions should be double-spaced, in Times 11-point or larger font, printed on one side of the page only, contain footnotes rather than endnotes, and have numbered pages. Microsoft Word format is preferred, but other formats also will be accepted. Electronic submissions can be sent to the Utah Law Review Expresso account or via e-mail at [e-mail], or a hard copy can be sent to: Executive Articles Editor, Utah Law Review, S.J. Quinney College of Law, University of Utah, 332 S 1400 E Rm. 101, Salt Lake City, UT 84112-0730. SUBSCRIPTIONS: The Utah Law Review is published quarterly. All subscriptions begin January 1 and expire on December 31. For subscriptions, renewals, or cancelations of the Utah Law Review, please send notice and/or payment to: Utah Law Review, Attn: Angela Turnbow, S.J. Quinney College of Law, University of Utah, 332 S 1400 E Rm. 101, Salt Lake City, UT 84112-0730. PRICES: $40.00 per volume per year; United States and possessions $45.00 per volume per year; foreign (postage included) To subscribe to more than one copy of the Utah Law Review, send notice of the number of copies requested along with payment. Please note that renewal payment must be received in advance before any issues are sent out to a subscriber. If you work with a subscription agency, send notice and payment to that agency and it will contact the Utah Law Review regarding the status of your subscription. The Utah Law Review can now be accessed online at: http://epubs.utah.edu/index. php/ulr. ISSUES AND REPRINTS: Single issues and back issues are available from William S. Hein & Co., Inc., 1285 Main Street, Buffalo, New York 14209-1987; telephone [phone]; e-mail: [e-mail]. Reprints are available from Joe Christensen, Inc., 1540 Adams Street, Lincoln, NE 68521; telephone [phone]; e-mail: [e-mail]. iii UTAH LAW REVIEW VOLUME 2012 NUMBER 1 S.J. QUINNEY COLLEGE OF LAW AT THE UNIVERSITY OF UTAH FACULTY & ADMINISTRATION 2011-2012 AARON, RICHARD I. (1966), Professor of Law Emeritus. A.B., 1959, Harvard Univ.; J.D., 1962, Univ. of Wisconsin. ADLER, ROBERT W. (1994), James I. Farr Chair in Law. B.A., 1977, John Hopkins Univ.; J.D., 1980, Georgetown Univ. AGUILAR, JR., REYES (1992), Associate Dean for Admission and Financial Aid. B.A., 1988, Texas A&M Univ.; J.D., 1992, Univ. of Utah. ANDERSON, JENSIE L. (1999), Clinical Professor of Law. B.F.A., 1985, J.D., 1993, Univ. of Utah. ANGHIE, ANTONY T. (1995), Samuel D. Thurman Professor of Law. B.A., 1986, LL.B., 1987, Monash Univ.; S.J.D., 1995, Harvard Univ. BOOHER, TROY L. (2011) Visiting Professor of Law. B.A., 1994, B.A., 1996, Metropolitan State College; J.D., 2002, Ph.D., 2010 Univ. of Utah. BROWN, TENEILLE R. (2009), Associate Professor of Law and Adjunct Professor of Internal Medicine. B.A., 2000, Univ. of Pennsylvania; J.D., 2004, Univ. of Michigan. CASSELL, PAUL G. (1991), Ronald N. Boyce Presidential Chair in Criminal Law. B.A., 1981, J.D., 1984, Stanford Univ. CHAHINE, KENNETH GREGORY (2009), Visiting Assistant Professor of Law. Ph.D., 1992, Univ. of Michigan; J.D., 1996, Univ. of Utah. CHIANG, EMILY (2009), Associate Professor of Law. B.A., 1998, Yale Univ.; J.D., 2001, Harvard Univ. CHODOSH, HIRAM E. (2006), Dean and Professor of Law. B.A., 1985, Wesleyan Univ.; J.D., 1990, Yale Univ. DAVIES, LINCOLN L. (2007), Associate Professor of Law. B.S., 1997, Univ. of Michigan; J.D., 2000, Stanford Univ. DICKEY, BARBARA J. (1999), Associate Dean for Student Affairs. B.A., 1975, M.A., 1980, J.D., 1987, Univ. of Utah. FIRMAGE, EDWIN B. (1966), Professor of Law Emeritus. B.S., 1958, Boston College; LL.B., 1961, Georgetown Univ.; S.J.D., 1967, Univ. of Michigan. FLORES, ROBERT L. (1990), Professor of Law. B.A., 1978, J.D., 1987, Univ. of Utah. FRANCIS, LESLIE P. (1982), Alfred C. Emery Professor of Law and Professor of Philosophy. B.A., 1967, Wellesley College; Ph.D., 1974, Univ. of Michigan; J.D., 1981, Univ. of Utah. GEORGE, ERIKA (2003), Professor of Law. B.A., 1992, M.A., 1993, Univ. of Chicago; J.D., 1996, Harvard Univ. GUIORA, AMOS N. (2007), Professor of Law. B.A., 1979, Kenyon College; J.D., 1985, Case Western Reserve Univ. HILL, DAVID S. (2003), Clinical Professor of Law. B.S., 1991, B.A., 1992, J.D., 2001, Univ. of Utah. HOLBROOK, JAMES R. (2003), Clinical Professor of Law. B.A., 1966, Grinnell College; M.A., 1968, Indiana Univ.; J.D., 1974, Univ. of Utah. iv JOHNSON, CHRISTIAN (2008), Professor of Law. B.A., 1984, M.Pr.A., 1985, Univ. of Utah; J.D., 1990, Columbia Univ. KEITER, ROBERT B. (1993), Wallace Stegner Professor of Law and Director, Wallace Stegner Center for Land, Resources, and the Environment. B.A., 1968, Washington Univ.; J.D., 1972, Northwestern Univ. KESSLER, LAURA T. (2001), Professor of Law. B.A., 1988, George Washington Univ.; J.D., 1993, Univ. of Maryland; LL.M., 1999, J.S.D., 2006, Columbia Univ. KINGSBURY, BENEDICT (2011), Visiting Professor of Law. L.L.B, 1981, Univ. of Canterbury, New Zealand; M. Phil., 1984, D.Phil., 1990, Oxford Univ. KOGAN, TERRY S. (1984), Professor of Law. B.A., 1971, Columbia College; B. Phil., 1973, Oxford Univ.; J.D., 1976, Yale Univ. LOCKHART, WILLIAM J. (1964), Professor of Law Emeritus. B.A., 1955, J.D., 1961, Univ. of Minnesota. LUND, THOMAS (1978), Professor of Law. B.A., 1964, Harvard Univ.; LL.B., 1967, Columbia Univ.; Ph.D., 1978, Oxford Univ. MABEY, RALPH R. (2007), Professor of Law. B.A., 1968, Univ. of Utah; J.D., 1972, Columbia Univ. MALLAT, CHIBLI W. (2007), Presidential Professor of Law. B.A., 1982, Beirut Univ. College; Ph.D., 1990, Univ. of London; LL.M., 1983, Georgetown Univ. MARTINEZ, JOHN (1984), Professor of Law. B.A., 1973, Occidental College; J.D., 1976, Columbia Univ. MATHESON, JR., SCOTT M. (1985), Hugh B. Brown Presidential Endowed Chair, Professor of Law. B.A., 1975, Stanford Univ.; M.A., Oxford Univ.; J.D., 1980, Yale Univ. MCCORMACK, WAYNE (1978), W. Wayne Thode Professor of Law. B.A., 1966, Stanford Univ.; J.D., 1969, Univ. of Texas. MCLAUGHLIN, NANCY ASSAF (2000), Associate Dean for Faculty Research and Development and Robert S. Swenson Professor of Law. B.S., 1987, Univ. of Massachusetts, Amherst; J.D., 1990, Univ. of Virginia. MEDWED, DANIEL S. (2004), Professor of Law. B.A., 1991, Yale Univ.; J.D., 1995, Harvard Univ. MITCHELL, BONNIE L. (1987), Clinical Professor of Law. B.A., 1981, J.D., 1984, Univ. of Utah. MORRIS, JOHN K. (1979), General Counsel, University of Utah and Professor of Law. B.A., 1966, Univ. of California at Los Angeles; J.D., 1969, Univ. of California at Berkeley. PETERSON, CHRISTOPHER (2008), Associate Dean for Academic Affairs and Professor of Law. B.S., 1997, H.B.A., 1997, J.D., 2001, Univ. of Utah. JAMIE PLEUNE (2011), Visiting Professor of Law. B.A., 2000, Colorado College; J.D., 2007, Univ. of Utah; L.L.M., 2011 Georgetown Univ. Law Center. POULTER, SUSAN R. (1990), Professor of Law Emerita. B.S., 1965, Ph.D., 1969, Univ. of California at Berkeley; J.D., 1983, Univ. of Utah. REITZE, JR., ARNOLD W. (2007), Professor of Law. B.A., 1960, Fairleigh Dickinson Univ.; J.D., 1962, Rutgers Univ.; M.P.H., 1986, John Hopkins Univ. REUSCH, RITA T. (1986), Director, James E. Faust Law Library and Professor of Law. B.A., 1972, Univ. of Minnesota; J.D., 1975, Univ. of Idaho; M.L.L., 1978, Univ. of Washington. RICHARDS,WILLIAM R. (1996), Clinical Professor of Law. B.A., 1981, J.D., 1984, Univ. of Utah. RINEHART, AMELIA (2010), Associate Professor of Law. B.S.E., 1996, M.S.E., 1997, Tulane Univ.; J.D., 2002, Univ. of Chicago. ROSKY, CLIFFORD J. (2008), Associate Professor of Law. B.A., 1996, Amherst College; J.D., 2001, Yale Univ. v SKIBINE, ALEXANDER TALLCHIEF (1989), S.J. Quinney Professor of Law. B.A., 1973, Tufts Univ.; J.D., 1976, Northwestern Univ. SMITH, LINDA F. (1984), Director, Clinical Program and Professor of Law. B.A., 1973, Ohio State Univ.; J.D., 1976, Yale Univ. TETER, MICHAEL (2011), Associate Professor of Law. B.A., 1999, Pomona College; J.D., 2002, Yale Univ. THREEDY, DEBORA L. (1986), Professor of Law. B.A., 1973, Beloit College; J.D., 1980, Loyola Univ. Chicago. WILDERMUTH, AMY J. (2003), Professor of Law. A.B., 1994, B.S., 1994, Washington Univ.; M.S., 1998, J.D., 1998, Univ. of Illinois. James E. Faust Law Library FACULTY CRAIGLE, VALERI (2008), Access Technologies Librarian. B.S., 1994, Univ. of Utah; M.L.S., 2005, Emporia State Univ. DARAIS, SUZANNE (1990), Head of Information Technical Services and Adjunct Professor of Law. B.S., 1983, Univ. of Utah; J.D., 1989, Brigham Young Univ.; M.L.S., 1990, Univ. of Washington. MCPHAIL, ROSS (2007), Assistant Librarian and Adjunct Professor of Law. B.A., 1990, Quincy College.; J.D., 1994, Hamline Univ. STEPHENSON, LINDA (1986), Head of Information Access Services and Adjunct Professor of Law. M.A., 1983, Univ. of Missouri at Columbia; J.D., 1987, Univ. of Missouri-Kansas City. WARTHEN, R. LEE (1985), Assistant Director, Head of Collection Building and Maintenance and Adjunct Professor of Law. B.A., 1973, M.L.S., 1975, M.A., 1983, J.D., 1984, Brigham Young Univ. vi UTAH LAW REVIEW VOLUME 2012 NUMBER 1 BOARD OF EDITORS 2011-2012 Managing Editor J.MASON KJAR Executive Footnote Editor THOMAS GOODWIN Note & Comment Editors KIMBERLY CHILD MATT DODD BRADFORD LIDDELL SARA MONTOYA Senior Staff JORDAN TOONE LAURIE EVANS ABBOTT DICK BALDWIN KIMBERLY BARNES RYAN BECKSTROM DANIEL BEZDJIAN CLARK COLLINGS DOUGLAS CRAPO TARYN EVANS ERIC A. OLSON TERRY S. KOGAN Editor-in-Chief JAVIER M.MIXCO Executive Articles Editor ERIKA SKOUGARD Symposium Editor MINDI HANSEN Associate Editors REBECCA HELD ALAN HOUSTON MEGHAN SHERIDAN KASSIDY WALLIN Staff Members JARED FORSGREN MELANIE GRAYSON RUTH HACKFORD-PEER LAUREN HANSEN SAMUEL JOHNSTON CYNTHIA LOVE NATHAN MARIGONI ADRIENNE MCKELVEY NATHANAEL MITCHELL Advisory Committee PATRICIA M. LEITH CLAYTON J. PARR Editorial Assistant ANGELA TURNBOW Executive Text Editor ALEX JACOBSON Executive Online Editor BARRY STRATFORD Articles Editors MAUREEN TOOMEY ARMOS RACHAEL BOETTCHER LAUREN HOSLER MATTHEW KAUFMANN ADAM PACE CHRISTOPHER PETERSEN KRISTIN SCHERRUBLE MARC THERRIEN DAVID MOOERS EMILY PATTEN TARA PINCOCK JOHN PLIMPTON KARINA SARGSIAN MARIANNE SCHUMANN BRIAN VOEKS TERA WOODS AMY J.WILDERMUTH MICHAEL D. ZIMMERMAN vii UTAH LAW REVIEW VOLUME 2012 NUMBER 1 TABLE OF CONTENTS ARTICLES Law, Fact, and Discretion in the Federal Courts: An Empirical Study Robert Anderson IV 1 Bargained Justice: Plea-Bargaining's Innocence Problem and the Brady Safety-Valve Lucian E. Dervan 51 Recantations Reconsidered: A New Framework for Righting Adam Heder & Wrongful Convictions Michael Goldsmith 99 Remedying the Misuse of Nature Sanne H. Knudsen 141 Due Process Denied: The Forgotten Constitutional Limits on Choice of Law in the Enforcement of Employee Covenants Not to Compete David A. Linehan 209 The Penguin and the Cartel: Rethinking Antitrust and Innovation Policy for the Age of Commercial Open Source Stephen M. Maurer 269 Class Action Defendants' New Lochnerism Mark Moller 319 A Minimalist Approach to Same-Sex Divorce: Respecting States that Permit Same-Sex Marriages and States that Refuse to Recognize Them Robert E. Rains 393 The Common Law of Disability Discrimination Mark C. Weber 429 viii NOTES E-Verify During a Period of Economic Recovery and High Unemployment Emily Patten 475 Incentive for Innovation or Invitation to Inhumanity?: A Human Rights Analysis of Gene Patenting and the Case of Myriad Genetics Laurie E. Abbott 497 1 LAW, FACT, AND DISCRETION IN THE FEDERAL COURTS: AN EMPIRICAL STUDY Robert Anderson IV* INTRODUCTION The standards of review that appellate courts use to review trial court decisions occupy an enigmatic position in legal practice and scholarly commentary.1 The conventional wisdom among appellate advocates and judges is that standards of review are key determinants-perhaps the key determinant-in the success of an appeal.2 Indeed, one prominent federal appellate judge wrote that the standard of review "more often than not determines the outcome" of an appeal3 and another described the standard of review as "everything."4 Flowing against this mainstream view, however, has long been an undercurrent of academic skepticism questioning whether standards of review actually constrain appellate judges.5 Even among appellate advocates, standards of review are sometimes viewed as "mere legalese"6 or "meaningless post hoc rationalizations" of appellate judicial decisions.7 Thus, the divergent streams of thinking mean that even though the * © 2012 Robert Anderson IV. Associate Professor of Law, Pepperdine University School of Law. J.D., NYU (2000); Ph.D., Stanford (2008). I would like to thank Emily Graves and Ryan Griffee for excellent research assistance. 1 The term "standards of review" is regularly used to refer to at least four different types of review: (1) appellate review of trial court decisions; (2) appellate review of jury decisions; (3) judicial review of administrative agency action; and (4) judicial review of government action under the Constitution (e.g., "strict scrutiny" or "rational basis" review). See 1 STEVEN A. CHILDRESS & MARTHA S. DAVIS, FEDERAL STANDARDS OF REVIEW § 1.03 (4th ed. 2010). This Article deals primarily with the first category-appellate review of trial court decision-making-or what the Supreme Court sometimes calls "court/court review." Dickinson v. Zurko, 527 U.S. 150, 153 (1999). 2 See, e.g., HARRY T. EDWARDS & LINDA A. ELLIOTT, FEDERAL COURTS STANDARDS OF REVIEW, at vii (2007) (stating that standards of review "are critically important in determining the parameters of appellate review"). 3 Patricia M. Wald, The Rhetoric of Results and the Results of Rhetoric: Judicial Writings, 62 U. CHI. L. REV. 1371, 1391 (1995). 4 EDWARDS & ELLIOTT, supra note 2, at vii (quoting Judge Deanell R. Tacha of the Tenth Circuit Court of Appeals). 5 John Stick, Can Nihilism Be Pragmatic?, 100 HARV. L. REV. 332, 362 n.130 (1986) (identifying standards of review as "notorious for being open to judicial manipulation" along with standing and procedural due process). 6 EDWARDS & ELLIOTT, supra note 2, at vii ("[W]e have been surprised by the number of attorneys who treat the standards of review as mere ‘legalese.'"). 7 CHILDRESS & DAVIS, supra note 1, at 1-14 (explaining it is "tempting" to view standards of review in this way, but arguing that they do "appear to have real value"). 2 UTAH LAW REVIEW [NO. 1 standard of review is the single most commonly invoked legal doctrine in appellate opinions,8 it is also probably among the least understood. The doctrinal perspective on standards of review is relatively easy to describe, at least in its broad outlines. The federal courts are divided into trial and appellate levels, each with a differentiated role in the adjudicatory process,9 and the standard of review largely delineates, at least in theory, the line between those roles.10 The standard of review divides judicial decisions into three principal categories-fact, law, and discretion-each allocated to the trial or appellate level.11 Trial courts have the primary responsibility for fact-finding and discretionary rulings on evidence and procedure.12 As a result, factual determinations and discretionary rulings are reviewed deferentially by the appellate courts13 and rarely reversed.14 The appellate courts, in contrast, have primary responsibility for interpretation and development of the law.15 As a result, legal conclusions of the trial court are freely (nondeferentially) reviewed by the appellate courts, and are more frequently reversed.16 In each case, standards of review delineate the boundaries between trial and appellate roles, determining the degree of deference appellate courts are required to give to trial courts. This textbook description of the adjudicatory process might be called the "legal model" of judicial decision-making.17 8 The standard of review is also typically among the first issues discussed in an appellate opinion. See Wald, supra note 3, at 1391 ("Courts talk about the standard of review up front in most opinions . . . ."). 9 Although the discussion in this Article is oriented primarily towards the federal courts of the United States, the division of roles between trial and appellate courts appears in most judicial systems around the world, although often with different divisions of roles. See HENRY J. ABRAHAM, THE JUDICIAL PROCESS 1 (1998). 10 Martin B. Louis, Allocating Adjudicative Decision Making Authority Between the Trial and Appellate Levels: A Unified View of the Scope of Review, the Judge/Jury Question, and Procedural Discretion, 64 N.C. L. REV. 993, 997 (1986) (arguing that the standard of review is "the principal means by which adjudicative decisional power and responsibility are divided between the trial and appellate levels"). 11 Pierce v. Underwood, 487 U.S. 552, 559 (1988). There is another category-mixed questions of law and fact-that is sometimes added as a category of its own, and sometimes decomposed into the categories of law and fact. See Maurice Rosenberg, Standards of Review, in RESTRUCTURING JUSTICE 32 (Arthur D. Hellman ed., 1990). 12 See Edward H. Cooper, Civil Rule 52(a): Rationing and Rationalizing the Resources of Appellate Review, 63 NOTRE DAME L. REV. 645, 657 (1988) ("[T]rial courts are primarily responsible for sifting the evidence and finding the facts . . . ."). 13 See infra Part I.A. 14 See infra notes 50-51 and accompanying text. 15 See Cooper, supra note 12, at 657 ("[A]ppellate courts are primarily responsible for developing the law."). 16 See infra notes 42-46 and accompanying text. 17 Of course, there is no single "legal model" of appellate decision-making. The term is used primarily by attitudinal scholars to refer to judicial decision-making constrained by legal principles. JEFFREY A. SEGAL & HAROLD J. SPAETH, THE SUPREME COURT AND THE ATTITUDINAL MODEL REVISITED 45-53 (2002). 2012] LAW, FACT, AND DISCRETION 3 There are reasons to doubt, however, whether the legal model's prescriptive ideal of trial and appellate roles matches the descriptive reality of appellate adjudication. The linchpin that holds the legal model together is the idea of deference-that an appellate court will voluntarily "set[] aside its own judgment" in favor of the judgment of the trial court on factual and discretionary issues.18 In other words, the idea of deference requires that an appellate court affirm at least some trial court decisions even when the appellate might have made a different decision had it originally decided the matter itself.19 This assumption becomes problematic as a practical matter, however, when the ideology of the trial court and the appellate court diverge. When the appellate court disagrees with a finding of fact or discretionary ruling made by a trial court it might be tempted to accord less deference to that decision than the legal model would prescribe. And even if the appellate court does defer to the trial court, that very deference might perversely create an incentive for the trial court to manipulate its own decision-making process, favoring rulings based on factual or discretionary grounds to insulate itself from reversal. The skeptic would argue, therefore, that the dual temptation to deviate from the norm of deference could undermine the legal model, distorting fact-finding and legal interpretation at both the trial and appellate levels. The theoretical concern that the formal legal model of law, fact, and discretion might not adequately capture the reality of the adjudicative process has been an important topic in legal theory for many decades. At least as far back as the legal realism movement,20 scholars expressed skepticism about the constraining force of legal doctrine and the accuracy of factual determinations. In 1930, Leon Green suggested that the fuzzy line between law and fact could invite manipulation by trial courts.21 In 1949, Jerome Frank outlined the problematic interaction of facts and legal rules in a perspective known as "fact skepticism," arguing that "trial-court fact-finding is the soft spot in the administration of justice."22 Green and Frank both noted that findings of fact, like conclusions of law, are malleable and subjective at the trial court level. At the appellate level, however, findings of fact are theoretically "fixed" and insulated from review. The appellate courts, for their part, were seen as manipulating the very definitions of law, fact, and discretion to recapture control over factual or discretionary questions insulated 18 Paul Horwitz, Three Faces of Deference, 83 NOTRE DAME L. REV. 1061, 1073 (2008). 19 Id. at 1074 (describing a "provisional definition of deference" as one decision-maker's willingness to follow another decision-maker's "determination, despite the fact that it might have reached a different conclusion had it reasoned independently"). 20 Legal realism was an early twentieth century legal philosophical movement that, in broad terms, explored the extent to which legal reasoning actually mattered in the arguments that lawyers made and the final decisions of judges. FREDERICK SCHAUER, THINKING LIKE A LAWYER: A NEW INTRODUCTION TO LEGAL REASONING 124-47 (2009). 21 See infra note 71 and accompanying text. 22 See, e.g., JEROME FRANK, COURTS ON TRIAL 74 (1949). 4 UTAH LAW REVIEW [NO. 1 from their review.23 Thus, for decades, scholars have argued that the malleability of facts at trial combined with their rigidity on appeal could create an important vulnerability in a hierarchical court system-encouraging manipulation of deferential standards of review by both trial and appellate courts. The realists and their contemporaries did not identify clearly, however, what empirical implications flowed from this vulnerability, the details of which have been sketched by attitudinal judicial politics scholarship and positive political theory literature in the last decade.24 In particular, one line of research in this literature has developed a theory that courts and administrative agencies try to further their ideological and policy goals through selecting "decision instruments" that make their decisions more difficult to overturn.25 The basic argument is that to the extent that reviewing courts or other political bodies find reversing certain types of decisions more costly than others, lower level strategic actors can protect their decisions from reversal by basing decisions on grounds that are more costly to overturn. If some types of decisions by a trial court-for example, factual determinations and legal rulings-are more costly for the appellate court to review, the trial court might have incentives to base decisions on grounds that are insulated from review, such as factual determinations. These contributions from positive political theory build on the skepticism of the legal realists to answer in part the realists' open question-the extent to which the malleability of law and fact and the strategic use of standards of review influence judicial decision-making. This Article builds on the foundation of the "decision instrument" literature to develop a theory of strategic interaction based on asymmetric information between the federal trial and appellate courts. This Article's theory suggests that although clear error and abuse of discretion are both deferential standards of review, they will have very different effects on interaction between trial and appellate courts.26 Specifically, the theory argues that clear error review, which applies to district court fact-finding, is associated with an informational advantage to the trial court 23 See Charles Alan Wright, The Doubtful Omniscience of Appellate Courts, 41 MINN. L. REV. 751, 778 (1957). 24 The "attitudinal" model is a political science approach to studying courts that "sees judicial decision-making as determined by the attitudes or preferences of individual judges, whose votes in particular cases reflect their sincere policy preferences largely unconstrained by legal precedent." Pauline T. Kim, Lower Court Discretion, 82 N.Y.U. L. REV. 383, 384-86 (2007). Positive political theory approaches are similar in their focus on judges' policy objectives, but positive political theory approaches "emphasize strategic interactions among judges and between judges and other political actors," where judges "tak[e] account of the likely response of other actors and the institutional context in which they operate." Id. at 384-85. 25 See Emerson H. Tiller & Pablo T. Spiller, Strategic Instruments: Legal Structure and Political Games in Administrative Law, 15 J.L. ECON. & ORG. 349, 350-52 (1999); Pablo T. Spiller & Matthew L. Spitzer, Judicial Choice of Legal Doctrines, 8 J.L. ECON. & ORG. 8, 10 (1992). 26 See infra notes 42-56 and accompanying text. 2012] LAW, FACT, AND DISCRETION 5 that will exacerbate ideological conflict between the trial and appellate courts.27 In contrast, the theory predicts that abuse of discretion review of procedural and evidentiary rulings is associated with relative symmetry of information that will lead to district court compliance and an absence of overt ideological conflict.28 The Article then gathers empirical evidence to examine whether the "legal model," the "decision instrument" theory, or the "informational" approach developed in this Article best explains the role of ideology in standards of review. The empirical analysis section uses newly developed data to peer inside the "black box" of judicial fact-finding and discretionary decisions using an analysis of appellate outcomes. The analysis extends the decision instrument literature by examining how two different deferential standards of review-clear error and abuse of discretion-affect the probability of reversal of district court decisions, and how those deferential review standards differ from one another and from de novo review. The answers to these basic questions lay the necessary foundation for understanding how deferential review might be used strategically. The analysis then considers how conflicting ideology of the district and appellate courts interacts with standards of review, specifically examining whether judges strategically invoke deferential review.29 Finally, the empirical analysis tests whether and how the effect of ideological divergence between the courts varies under conditions of informational asymmetries between the trial and appellate courts.30 The results support the conclusion that standards of review matter, but not in the way that either the conventional legal model of deference or the positive political theory would predict.31 The effect of deferential review appears to be more complex, in part because, as the theory developed in this Article suggests, the two principal deferential review standards (clear error and abuse of discretion) have qualitatively different effects.32 The results are consistent with the legal model in that deferential standards of review appear to considerably decrease the probability of outright reversal, and the analysis reveals no evidence that judges manipulate standards of review. On the other hand, the results suggest the counterintuitive conclusion, predicted by the informational theory, that findings of fact are associated with more manifested ideological disagreement than discretionary rulings or conclusions of law. The Article concludes that trial judges rely on asymmetric information inherent in fact-finding to achieve their goals when the clear error standard already applies, rather than invoking that standard strategically to insulate their decisions from review.33 Appellate judges, in turn, use ideological cues to discipline these attempts to capitalize on the information 27 See infra notes 98-99 and accompanying text. 28 See infra notes 203-204 and accompanying text. 29 See infra Part III.B. 30 See infra Part III.C. 31 See infra Part IV. 32 See infra Part III.C. 33 See infra Part IV.F.1. 6 UTAH LAW REVIEW [NO. 1 asymmetry, making fact-finding more ideologically contentious than legal conclusions.34 In Part I, the Article briefly describes the doctrinal framework of standards of review and sketches a theory of how law, fact, and discretion interact in a judicial hierarchy. The theory suggests three fundamental, but unanswered research questions for empirical testing: (1) whether standards of review affect the probability of reversal of the district court, (2) whether judges select ideologically among deferential and nondeferential standards of review, and (3) how the effect of ideology varies among the standards of review. Part II describes the data and measurement strategy used in this Article, including original data on standards of review developed for this analysis. Part III presents the results of the empirical analysis using the new dataset of standards of review to examine these research questions. The results suggest a complex relationship among standards of review, ideology, and information that does not fully correspond with either the legal model or the positive political theory approaches to the judiciary. Part IV interprets the results and discusses implications for policy decisions about the courts of appeals. I. DOCTRINE, DEFERENCE AND INFORMATION A. Legal Doctrine The organization of the federal judiciary is premised on the division of labor between trial and appellate courts, the boundaries of which are delineated by the standards of review. The standard of review divides trial court decisions into three broad categories, each of which is subject to a different standard: (1) conclusions of law, (2) fact-finding, and (3) discretionary rulings.35 Trial court conclusions of law, or what Hart and Sacks called the "law declaration" function,36 specify what consequences the legal doctrine attaches to various factual situations, including the factual situation as found by the trial court.37 Trial court fact-finding is the process 34 See infra Part IV.F.2. 35 See Pierce v. Underwood, 487 U.S. 552, 559 (1988). See also Professor Rosenberg's memorable description: "[A]ll appellate Gaul is divided into three parts for review purposes: questions of fact, of law and of discretion." Maurice Rosenberg, Judicial Discretion of the Trial Court, Viewed From Above, 22 SYRACUSE L. REV. 635, 645-46 (1971). In later work, Professor Rosenberg added mixed questions of law and fact for a total of four categories. Rosenberg, supra note 11, at 32. Edwards and Elliott identify four principal standards of review-de novo, clearly erroneous, abuse of discretion, and plain error. EDWARDS & ELLIOTT, supra note 2, at 3. 36 HENRY M. HART, JR. & ALBERT M. SACKS, THE LEGAL PROCESS: BASIC PROBLEMS IN THE MAKING AND APPLICATION OF LAW 350 (William N. Eskridge, Jr. & Philip P. Frickey eds., 1994). 37 The devotee of Hart and Sacks will note the absence of the third category of "law application" where courts apply the law to the facts so as to render a judgment or decision that binds the parties to a particular legal dispute. Id. at 351. This law application function 2012] LAW, FACT, AND DISCRETION 7 of determining the historical facts relevant to the case,38 or what has been called "a case-specific inquiry into what happened here."39 Discretionary rulings typically include trial court decisions on detailed questions of trial supervision, procedure, and evidence.40 Each of these basic components of the adjudicative process-law, fact, and discretion-is subject to its own level of appellate scrutiny according to a "standard of review."41 The decisions in the first category-questions of law-are reviewed "de novo" by the appellate court, meaning that the appellate court is not required or expected to give any deference to the trial court.42 Instead, the appellate court exercises its own judgment on the legal questions presented, exercising a form of review sometimes referred to as "free, independent, or even plenary review."43 In such cases, "[t]he appellate courts merely ask themselves whether they agree with the trial judge's resolution of the legal issue. If not, they reverse him quick as a flash . . . ."44 The appellate court is entitled to-and should-take into account the legal analysis of the trial court,45 but the appellate court has the right and even the duty to exercise its own independent judgment.46 In contrast to legal conclusions, factual determinations by the trial court are reviewed deferentially.47 Appellate review of factual findings by the district court is limited to whether those findings are "clearly erroneous,"48 a very high bar to is often assimilated into the "mixed question of law and fact." George C. Christie, Judicial Review of Findings of Fact, 87 NW. U. L. REV. 14, 53-54 (1992). 38 This is what Hart and Sacks call the "fact identification" function. HART & SACKS, supra note 36, at 350. 39 Henry P. Monaghan, Constitutional Fact Review, 85 COLUM. L. REV. 229, 235 (1985) (citing HART & SACKS, THE LEGAL PROCESS: BASIC PROBLEMS IN THE MAKING AND APPLICATION OF LAW 375 (tent. ed. 1958)). 40 See 1 PAUL G. ULRICH ET AL., FEDERAL APPELLATE PRACTICE: NINTH CIRCUIT § 4:6 (2d ed. 1999). 41 CHILDRESS & DAVIS, supra note 1, at 1-3. In addition to the degree of deference, the standard of review can also determine the materials the appellate court will look at in the course of its review, and even how hard the appellate court will look for error. Id. 42 See BLACK'S LAW DICTIONARY 924 (9th ed. 2009). 43 CHILDRESS & DAVIS, supra note 1, at 2-89. 44 Rosenberg, supra note 35, at 646. 45 See Salve Regina Coll. v. Russell, 499 U.S. 225, 232-33 (1991). 46 Id. at 238 ("When de novo review is compelled, no form of appellate deference is acceptable."). 47 There are exceptions to the deferential review of factual findings, of which the "constitutional fact" one is the most notable. See, e.g., Monaghan, supra note 39, at 231. This doctrine grew out of the "jurisdictional fact" doctrine of earlier cases. See Christie, supra note 37, at 26-31. 48 The standard of review for bench trial findings of fact in civil cases and criminal cases as to nonguilt questions is "clear error." CHILDRESS & DAVIS, supra note 1, § 7.01, at 7-3. The standard of review for jury findings of fact in both criminal and civil cases are variously described as "substantial evidence" or "reasonableness." Id. §§ 3.04, 7.01, at 7-3. 8 UTAH LAW REVIEW [NO. 1 reverse a trial judge's factual findings. The clear error standard does not require reversal only if there is no evidence to support the finding. Instead, as the canonical formulation of the clearly erroneous standard suggests, a finding of fact is clearly erroneous "when although there is evidence to support it, the reviewing court on the entire evidence is left with the definite and firm conviction that a mistake has been committed."49 The key point is that the standard of review constrains the appellate court to defer to the trial court's findings of fact, meaning that the appeal is primarily an appeal of legal issues, not a new trial of the whole case. Indeed, the conventional wisdom on review of findings of fact is very simple: "appellate courts do not engage in factual evaluation,"50 so that the clear error review is "usually cited to justify refusal to interfere with the fact findings made in the trial court."51 The third and final major standard of review-abuse of discretion-also requires deference to the trial court.52 The abuse of discretion standard applies to a wide range of procedural and evidentiary decisions that are entrusted to the discretion of the trial judge.53 Consistent with the broad range of decisions that are considered discretionary, there is no one formulation of the abuse of discretion standard.54 Instead, a highly contextual inquiry varies depending on the setting. The concept of discretion in this context "implies the power to choose within a range of acceptable options."55 The trial judge can be reversed when he or she goes beyond that acceptable range, but the appellate court must defer to a decision within the acceptable range, even if the appellate court would have made a The standard of review for judicial findings of fact in nonguilt issues is the same. Id. § 7.01, at 7-3. 49 United States v. U.S. Gypsum Co., 333 U.S. 364, 395 (1948). The "substantial evidence" standard applied to jury findings of fact is in theory even more deferential, although in practice many courts tend to equate the two standards. CHILDRESS & DAVIS, supra note 1, § 2.07, at 2-43 to 2-46. 50 Chad M. Oldfather, Appellate Courts, Historical Facts, and the Civil-Criminal Distinction, 57 VAND. L. REV. 437, 438 (2004). 51 CHARLES ALAN WRIGHT, THE LAW OF FEDERAL COURTS § 96, at 647 (4th ed. 1983). 52 See Gen. Elec. Co. v. Joiner, 522 U.S. 136, 143 (1997) (reprimanding the circuit court for failing "to give the trial court the deference that is the hallmark of abuse-of-discretion review"). 53 For example, rulings about the probative versus the prejudicial effect of evidence under Rule 403 of the Federal Rules of Evidence is subject to abuse of discretion review, see Old Chief v. United States, 519 U.S. 172, 183 n.7 (1997), as is the decision to grant or deny permanent injunctive relief, see Weinberger v. Romero-Barcelo, 456 U.S. 305, 320 (1982), and preliminary injunctions, see Ashcroft v. ACLU, 542 U.S. 656, 664 (2004), and the decision to admit or exclude expert testimony, see Joiner, 522 U.S. at 138-39. 54 See EDWARDS & ELLIOTT, supra note 2, at 67 ("[T]he variety of matters committed to the discretion of district judges means that the standard is necessarily variable. It implies no single level of scrutiny by the appellate courts."). See also CHILDRESS & DAVIS, supra note 1, § 4.01, at 4-12 ("[T]here is no such thing as one abuse of discretion standard."). 55 CHILDRESS & DAVIS, supra note 1, § 4.01, at 4-3. 2012] LAW, FACT, AND DISCRETION 9 different decision. The precise formulation of the standard will vary from case to case, but in all cases it is "deference that is the hallmark of abuse-of-discretion review."56 B. Hypotheses The legal doctrine of standards of review is clear: appellate courts are to defer to trial court findings of fact and discretionary rulings, but not to conclusions of law. In light of the skeptical undercurrent about standards of review, however, the question remains as to what causal effect, if any, the doctrinal command of deference actually has on the appellate court's review. As a preliminary matter, there is little doubt that appellate courts regularly defer to trial court decisions regardless of the standard of review, if for no other reason than the sheer volume of appellate disputes that make plenary review of all decisions impossible.57 Factual disputes in particular are especially complicated, and therefore would naturally receive some deference, whether or not the standard of review commanded such deference. But the conventional account of appellate process makes a much stronger claim-that appellate judges defer to trial court fact-finding and discretionary rulings not merely out of necessity, but also out of obedience.58 In other words, the conventional legal model claims it is the formal legal standard of review, not merely constraints on the appellate court's resources, that causes the appellate judge to review the trial court more deferentially. The conventional account is vulnerable, however, to at least three types of temptations that could undermine the prescriptive rule of deference. First, appellate judges might invoke deferential standards of review in their opinions and yet freely reverse decisions with which they disagree. Second, even if appellate judges faithfully defer when they apply deferential standards of review, appellate judges might strategically invoke deferential standards of review when they agree with the lower court's decision, and nondeferential standards when they disagree with the lower court's decision. Third, even if appellate judges faithfully defer according to the deferential standards and do not strategically invoke them, the very fact of deference may create an incentive for the trial courts to manipulate their decisions toward deferentially reviewed grounds of decision. It is possible that judges do not succumb to the three temptations and do not strategically choose between deferential and nondeferential standards. Indeed, deferential review may be a 56 Joiner, 522 U.S. at 143. 57 Indeed, the "legal model" may be inaccurate not only in that judges do not defer when they are required to, as explored in this Article, but in that judges do defer when they are not required to, whether out of sheer necessity or for other reasons. There is some evidence for this behavior, which is referred to as the "affirmance effect." See Chris Guthrie & Tracey E. George, The Futility of Appeal: Disciplinary Insights into the ‘Affirmance Effect' on the United States Courts of Appeals, 32 FLA. ST. U. L. REV. 357, 358 (2005); see also Kevin M. Clermont & Theodore Eisenberg, Litigation Realities, 88 CORNELL L. REV. 119, 150 (2002) (showing the "affirmance rate to be about 80%"). 58 See Clermont & Eisenberg, supra note 57, at 150-52. 1 0 UTAH LAW REVIEW [NO. 1 rational organizational institution that leaves room for ideological disagreement in the judicial hierarchy. The next three sections theorize the temptation possibilities and take preliminary steps toward understanding which account best describes the effect of deferential review. 1. Do Standards of Review Affect Reversal Rates? The first research question developed in this Subpart is whether deferential review is associated with a lower probability of reversal compared to de novo review. The general understanding of clear error review and abuse of discretion review is that reversal is very unlikely, tending to shift power from the appellate court to the trial court.59 Yet the question of what effect, if any, standards of review have on the outcome of individual cases has not been subject to extensive empirical analysis. The conventional wisdom must be tested against outcomes, because even if appellate panels invoke deferential standards of review when legal doctrine suggests they should, the judges may simply fail to actually defer to the trial court.60 In other words, the appellate court might recite the standard of review but not feel meaningful constraint on its decision to affirm or reverse. This is probably the most fundamental question to ask about standards of review, because if the standard of review is unrelated to the probability of reversal, then both the legal model and the "decision instrument" model are based on false assumptions.61 Why would the standard of review be viewed as a constraint the appellate panel might want to ignore? The most obvious reason would suggest that ideological disagreement between the trial and appellate courts would provide a strong temptation to ignore deferential standards of review.62 The first research question investigated, therefore, is whether standards of review or differences in ideology between the district court and appellate court better explain appellate outcomes. This involves comparing outcomes along two dimensions. First, as the 59 Louis, supra note 10, at 1045 ("Designating a procedural determination as discretionary results in a limited scope of review on appeal, few reversals, a reduced number of appeals, and, therefore, trial level hegemony over the question."). 60 Of course, the appellate court might then be subject to reversal by the circuit en banc or by the Supreme Court, which does occur from time to time. See, e.g., Gall v. United States, 552 U.S. 38, 56 (2007) ("[A]lthough the Court of Appeals correctly stated that the appropriate standard of review was abuse of discretion, it engaged in an analysis that more closely resembled de novo review of the facts presented and determined that, in its view, the degree of variance was not warranted."). 61 The legal model would be wrong because it commands deference to the trial court, which should decrease the reversal rate. The strategic model would be wrong because judges cannot strategically invoke standards of review if they make no difference to the outcome. 62 See, e.g., Guthrie & George, supra note 57, at 364-71 ("The Court is more likely to grant certiorari to review lower court decisions that are ideologically inconsistent with the Court's current majority because the Supreme Court acts ideologically, and thus it is more likely to grant certiorari to reverse than to affirm the lower court."). 2012] LAW, FACT, AND DISCRETION 11 standard of review changes does the rate of reversal or affirmance change? Second, as the ideology of the median member of the appellate panel diverges from the ideology of the district court judge, how does the probability of reversal change? The first hypothesized factor-the standard of review-although relatively straightforward and intuitive, does not appear to have been tested empirically in the court/court context until very recently.63 The second factor-the effect of ideology-does not appear to ever have been compared with standards of review.64 2. Do Judges Strategically Choose Standards of Review? The second research question is whether trial courts, appellate courts, or both strategically manipulate deferential review to achieve desired outcomes. If appellate courts do defer when deferential standards apply, as even some skeptical observers have conceded they likely do,65 trial courts might have an incentive to behave strategically. Specifically, a trial court might attempt to base a vulnerable decision on factual or discretionary grounds when the appellate court is likely to disagree with the trial court's decision, and on legal grounds when the appellate court is likely to agree with the trial court's conclusion. Similarly, appellate panels might characterize favored trial court decisions as "factual" or "discretionary," invoking deferential review, while characterizing disfavored decisions as "legal," invoking de novo review.66 The necessary conditions for this type of manipulation are that: (1) deferential review affects the reverse/affirm decision and (2) deferential review is, at least in part, the object of choice by the trial court, the 63 See Corey Rayburn Yung, Judged by the Company You Keep: An Empirical Study of the Ideologies of Judges on the United States Courts of Appeals, 51 B.C. L. REV. 1133, 1136 (2010); see also Joshua B. Fischman & Max M. Schanzenbach, Do Standards of Review Matter? The Case of Federal Criminal Sentencing, 40 J. LEGAL STUD. 405, 406 (2011) ("Although standards of review are often the subject of intense political and judicial debate, there is little empirical evidence that these standards have an effect on either district or circuit judges."). Another important study examined the standards of review in the context of review of administrative decisions, but did not test the effect of ideology on court/court review. See FRANK B. CROSS, DECISION MAKING IN THE U.S. COURT OF APPEALS 53 (2007). 64 See, e.g., Yung, supra note 63, at 1136 ("Only recently have studies begun to develop comprehensive measures of ideology based upon actual judicial performance."). 65 Wright, supra note 23, at 770 ("The courts which have disregarded Rule 52 in substituting their judgment for that of the trial court could accomplish the same purpose while complying with the rule merely by announcing that the finding with which they disagree is ‘clearly erroneous.' But I think we can safely assume that appellate judges do make a conscientious attempt to confine their review to that authorized by law, and that, so far as human frailties permit, they do not regard a finding as clearly erroneous merely because it differs from the finding they might themselves have made."). 66 See, e.g., Henry J. Friendly, Indiscretion about Discretion, 31 EMORY L.J. 747, 776 (1982) ("A favorite method used by appellate courts to avoid the discretion rule in temporary injunction cases is to find that the district court proceeded on an erroneous view of the law . . . ."). 1 2 UTAH LAW REVIEW [NO. 1 appellate court, or both. The first condition-that standards of review matter-was described above.67 The second condition-the manipulability of deferential review-is described below.68 Do trial judges, appellate judges, or both have the flexibility to move freely between deferential review and nondeferential review, as some of the strategic instrument literature would suggest? The argument draws considerable support from the vast literature on the elusiveness of a distinction between law and fact.69 Indeed, the idea that judges think strategically about law and fact has been a part of the legal literature at least since the legal realist movement, and is still vibrant today.70 In a quotation from over 80 years ago, the legal realist Leon Green incisively remarked: No two terms of legal science have rendered better service than "law" and "fact." They are basic assumptions; irreducible minimums and the most comprehensive maximums at the same instant. They readily accommodate themselves to any meaning we desire to give them. In them and their kind a science of law finds its strength and durability. They are the creations of centuries. What judge has not found refuge in them? The man who could succeed in defining them would be a public enemy.71 The basic idea of the quotation is clear; the distinction between the terms "law" and "fact" is malleable at best, and the malleability is not entirely unwelcome to the judges who wield the terms. But Green goes farther, arguing that judicial actors-primarily trial judges in this context-can strategically insulate 67 See supra notes 31-34 and accompanying text. 68 See infra notes 76-81 and accompanying text. 69 See, e.g., Ronald J. Allen & Michael S. Pardo, Essay: The Myth of the Law-Fact Distinction, 97 NW. U. L. REV. 1769, 1771-90 (2003) (discussing the "legal doctrine surrounding the law-fact distinction"); Richard D. Friedman, Standards of Persuasion and the Distinction Between Fact and Law, 86 NW. U. L. REV. 916, 917-25 (1992) (discussing the law and fact distinction); Henry P. Monaghan, Constitutional Fact Review, 85 COLUM. L. REV. 229, 232-39 (1985) (discussing the "distinction between ‘questions of law' and ‘questions of fact'"). Skepticism about the distinction between law and fact goes back many decades, one of its forceful early proponents being Leon Green. See, e.g., LEON GREEN, JUDGE AND JURY 270 (1930) ("Such among others are those equally expansible and collapsible terms ‘law' and ‘fact.'"). 70 See, e.g., Bryan L. Adamson, Federal Rule of Civil Procedure 52(a) As An Ideological Weapon?, 34 FLA. ST. U. L. REV. 1025, 1028 (2007) ("If an appellate court wants to reweigh the facts as found by a lower court, it may characterize the trial court's factual findings as ‘legal conclusions' or ‘mixed questions of law and fact.'"); Christopher M. Pietruszkiewicz, Economic Substance and the Standard of Review, 60 ALA. L. REV. 339, 355-60 (2009) ("[A]n appellate court can easily review and decide a factual issue by simply recasting it as a question of law and applying a de novo standard of review instead of the more stringent clearly erroneous standard."). 71 GREEN, supra note 69, at 270. 2012] LAW, FACT, AND DISCRETION 13 their decisions from scrutiny and "take refuge" by choosing to ground the decision in law or in fact.72 The malleability of the law-fact distinction works both ways, however, creating strategic opportunities for both the reviewer and the reviewed. Appellate judges might tend to characterize decisions they wish to reverse as conclusions of law, and decisions they wish to affirm as findings of fact, selectively using de novo and deferential review standards as "ideological weapon[s]."73 After all, as Judge Wald has pointed out, appellate courts do not simply "apply ‘the law' to the facts," which would assume a clean distinction between the two, but they also "have to decide what the ‘standard of review' is,"74 meaning that the distinction is endogenous to the choice process. The "mixed question of law and fact" is the most obvious context in which the appellate court has an explicit opportunity to choose the standard of review. This is where many scholars observe the "struggle for power" between trial and appellate courts.75 Over the years, authors have argued that appellate courts have in fact manipulated the standard of review, often by moving more and more decisions from deferential review into the de novo category.76 The standard of review, therefore, is potentially a strategic choice for the appellate court and there is the possibility that courts manipulate standards of review just as Green's observed decades ago. This perspective has been invigorated in recent decades by positive political theory accounts of judicial decision-making.77 Here, the emphasis is typically on how the lower court or an administrative agency can strategically choose grounds for the decision that makes review by the higher-level actor more costly.78 This might include, for example, the choice of an agency to make policy by rulemaking or adjudication,79 or an appellate court's choice between critiquing an agency's statutory interpretation and critiquing an agency's reasoning process.80 In each case, the first choice (rulemaking and statutory interpretation) involves legal questions that can be 72 Indeed, trial judges can even manipulate the type of facts they rely on, shifting the emphasis from "ultimate facts" to "historical facts." See Louis, supra note 10, at 1015 ("Trial judges can still slant or stack their historical fact-findings to shield their ultimate fact-findings from reversal."). 73 See Adamson, supra note 70, at 1028-32. 74 Wald, supra note 3, at 1391. 75 Louis, supra note 10, at 1003 ("[T]heir inevitable struggle for power must center on the process of law application and in particular on the scope of review of mixed law/fact questions."). 76 See, e.g., Martin Shapiro, Appeal, 14 LAW & SOC'Y REV. 629, 648 (1980) ("Appellate courts have worked so hard at eroding the boundary between facts and law that today almost any issue can be characterized as a question of law or a mixed question of law and fact, and in either case appropriate for appellate consideration."). 77 See infra Part IV. 78 See infra Part IV. 79 Tiller & Spiller, supra note 25, at 349-51. 80 See Joseph L. Smith & Emerson H. Tiller, The Strategy of Judging: Evidence from Administrative Law, 31 J. LEGAL STUD. 61, 64-67 (2002). 1 4 UTAH LAW REVIEW [NO. 1 reviewed at low cost and the second choice (adjudication and reasoning process) involves fact-intensive inquiries that can be reviewed only at high cost. A few empirical studies have investigated these theories and whether their model predictions are supported.81 The work described above has dealt primarily with judicial-administrative interaction, but a more recent article in this line of research directly addresses the court/court strategic issues that this Article attempts to untangle.82 One recent piece, for example, specifically examines the district court's attempt to strategically characterize decisions as findings of fact or conclusions of law.83 It tests the hypothesis that district court judges will be more likely to ground their sentencing decisions as "departures" (law-based) rather than "adjustments" (fact-based) when there is alignment with the reviewing circuit court.84 In contrast, the theory argues that lower courts will be limited to the fact-based adjustments when there is ideological divergence between the sentencing court and the appellate court, because fact-based adjustments receive deferential review while law-based departures do not.85 If this is true and appellate judges routinely choose the standard of review based on their attitude toward the case before them, this might bias the results in the first two hypotheses. To test whether deferential review is strategically invoked, this Article examines the relationship between ideology and standards of review at the district court and appellate court level. The first part of the hypothesis is that appellate courts would apply more deferential standards in cases where they are ideologically close to the lower court, and more searching standards where they are ideologically farther from the lower courts. The second part of the hypothesis is that district courts will "take refuge" in deferential review standards when the district court is ideologically farther from the center of the relevant appellate circuit. If indeed the distinction between law and fact or between law and discretion is elusive, traditional legal theory does not offer a solution for dealing with this strategic behavior. 3. Do Judges Strategically Choose Within Standards of Review? In addition to the possibility that judges might ignore deferential review or strategically select among standards of review, judges might also behave ideologically within standards of review. In other words, judges might behave more ideologically under deferential review than under de novo review or vice versa. There are at least two types of reasons why judicial behavior might vary 81 See, e.g., id. at 64-67 (applying empirical methodologies to a study of judicial review of administrative action). 82 Max B. Schanzenbach & Emerson H. Tiller, Strategic Judging Under the U.S. Sentencing Guidelines: Positive Political Theory and Evidence, 23 J.L. ECON & ORG. 24, 24, 45-52 (2007) (applying a similar analysis in the context of sentencing). 83 Id. 84 Id. at 25-26. 85 See id. at 33. 2012] LAW, FACT, AND DISCRETION 15 under different standards of review. The first type of reason is that the command of deferential or nondeferential review itself might affect ideological behavior. One might hypothesize, for example, that appellate judges behave less ideologically when they are instructed to review the trial court deferentially. The second type of reason is that the context in which the relevant standard of review is applied might affect ideological behavior. One might hypothesize, for example, that clear error or abuse of discretion review might prove less ideological not because of the command of deference, but because of the narrow, limited nature of rulings on these issues that primarily affect the case before the court. To understand how the command and the context might affect decision-making within each standard of review, it is valuable to examine the purposes and applicability of deferential review. The basic test for whether deferential review is appropriate is remarkably functional-focusing on two questions: (1) whether the district court is "better positioned" to decide the issue and (2) whether "probing appellate scrutiny" will "contribute to the clarity of legal doctrine." 86 These two functional considerations flow from the two primary purposes of the appellate courts, error correction and development of a consistent, coherent body of law.87 The two inquiries are sometimes described collectively in terms of whether the issue is "fact-intensive," "fact-dependent" or "fact-bound,"88 but the inquiries are analytically separate, as the Supreme Court itself has recognized.89 The first type of situation where deference is appropriate is where the facts that are relevant to the decision are more accessible to the trial judge than to the appellate panel.90 This is the rationale for the deferential "clearly erroneous" 86 Miller v. Fenton, 474 U.S. 104, 114 (1985); see also Salve Regina Coll. v. Russell, 499 U.S. 225, 233 (1991). 87 See, e.g., Louis, supra note 10, at 1006 ("Appellate courts exercise two principal, overlapping functions: they declare the law and they supervise the results of lower level adjudication in individual cases."). 88 Cooter & Gell v. Hartmarx Corp., 496 U.S. 384, 403-05 (1990). 89 See, e.g., Pierce v. Underwood, 487 U.S. 552, 559-62 (1988). In this case, the Court explained the first reason as "the district court may have insights not conveyed by the record, into such matters as whether particular evidence was worthy of being relied upon, or whether critical facts could easily have been verified by the Government. Moreover, even where the district judge's full knowledge of the factual setting can be acquired by the appellate court, that acquisition will often come at unusual expense" and also the "impracticality of formulating a rule" when "multifarious and novel question, little susceptible, for the time being at least, of useful generalization . . . ." Id. at 560-62 (citations omitted). For the second reason, the court found that "[o]ne of the ‘good' reasons for conferring discretion on the trial judge is the sheer impracticability of formulating a rule of decision for the matter in issue." Id. at 561 (quoting Rosenberg, supra note 35, at 662- 63). 90 In some cases, such as where fact-finding depends heavily on oral testimony, the trial judge's advantage is obvious. But Rule 52(a) does not limit deference to such situations. See FED. R. CIV. P. 52(a). Indeed the only significant amendment to 52(a) occurred in 1985 to clarify that the Rule applies whether to "oral or documentary 1 6 UTAH LAW REVIEW [NO. 1 standard on review of facts themselves, but also applies to mixed questions of law and fact where "the district court may have insights not conveyed by the record," meaning that some facts "may be known only to the district court."91 The trial judge spends much more time interacting with the evidence, witnesses, and details of the case, and of course the trial judge alone has access to demeanor evidence, whereas the appellate court only has the "cold printed record."92 Indeed, in a very real sense, the trial judge is himself or herself a "witness[ ] to what transpires in the courtroom."93 As a result, facts may be completely lost on appeal,94 for as the Supreme Court has stated, "nervousness cannot be shown from a cold transcript."95 In these situations, the decision is "fact-bound" in a sense, but the reason for deference is really that the district court has information the appellate court cannot acquire without a tremendous investment of resources, if at all. The second inquiry, which examines whether de novo appellate examination will contribute to legal doctrine, is the one that focuses on whether the issue is fact-sensitive, fact-intensive, or fact-dependent. The rule is highly dependent on the precise facts of the case, whether or not those facts are peculiarly within the special command of the trial judge. The question is one of judicial resources, but here the appellate court is deciding whether formulating detailed rules on a particular issue exceeds the benefit from those rules.96 In such cases, deference is indicated when the issue is not appropriate for appellate decision because "impracticability of formulating a rule of decision" when the issue involves "multifarious, fleeting, special, narrow facts that utterly resist generalization . . . ."97 The appellate court may well have access to the same information as the trial court at relatively low evidence," see Cooper, supra note 12, at 648, as some courts had made a distinction between the two, id. at 651. The Supreme Court had already made the same clarification. Anderson v. City of Bessemer, 470 U.S. 564, 573-76 (1985). 91 Pierce, 487 U.S. at 560. 92 The earliest example of this heavily used term that I have been able to locate is the concurring opinion of Judge McFarland in People v. Phelan, 56 P. 424, 431-32 (1899) ("I desire to say, however, that if the testimony were to be considered simply as it stands in the cold, printed record, without any reference to the appearance, manner of testifying, etc., of the witnesses, I would not feel sure in holding that it warranted a conviction. But the credibility of the witnesses rests with the jury, and, as it depends upon many things which cannot be reproduced here, I do not feel that this court would be justified in setting aside the verdict on the ground of want of evidence."). 93 Christie, supra note 37, at 46 (citing JEROME FRANK, LAW AND THE MODERN MIND 109-10 (1930)). 94 Id. at 50 ("No record on appeal, however complete, can capture the richness of the trial situation."). 95 Snyder v. Louisiana, 552 U.S. 472, 479 (2008) (quoting Louisiana v. Snyder, 942 So. 2d 484, 496 (2006)). 96 See, e.g., RICHARD A. POSNER, ECONOMIC ANALYSIS OF LAW 586-90 (6th ed. 2003) (describing the cost-benefit tradeoff of detailed rules versus general standards). 97 Pierce v. Underwood, 487 U.S. 552, 561-62 (1988) (quoting Rosenberg, supra note 35, at 662-63). 2012] LAW, FACT, AND DISCRETION 17 cost, but no general rule could be formulated that would helpfully resolve future cases. The first type of argument for deferential review involves the high cost of the appellate court becoming informed of the facts, requiring a mechanism for the trial court to develop and communicate factual information to the appellate court. The paradigmatic example of this situation is the clear error review of judicial findings of fact. The second type of argument involves the high cost of communicating detailed rules from the appellate court to the trial court in fact-dependent contexts. The paradigmatic example of this situation is the abuse of discretion review of trial court rulings, such as evidence and procedure. In both cases, deference to the trial court is appropriate, but only in the first case does the trial court have private information. The distinct nature of these two reasons for deferential review leads to a third hypothesis about standards of review. If the district court has private information about the case it can behave in an ideological manner without perfect monitoring by the appellate court. That is, the appellate court cannot completely discipline ideological behavior by the trial court, because even if the case is appealed, the appellate court cannot invest the resources necessary to critically evaluate findings of fact. Because the appellate court cannot directly observe the underlying facts of the case, the appellate court may feel compelled to rely on signals, such as the trial court's ideology, to infer the "true" facts. The appellate court reviewing a decision of a district court judge known to have a different ideological viewpoint might invest more resources in reviewing the facts underlying the decision. For abuse of discretion review of evidentiary and procedural decisions, however, the trial court is not likely to have a considerable informational advantage relative to the appellate court. The appellate court can generally inform itself of all relevant information about the trial court's procedural and evidentiary rulings at relatively low cost.98 The district court knows this, which means that the district court would rationally hew much closer to the appellate panel's likely preferences (i.e., the ideological makeup of the whole circuit from which panels are drawn). In other words, the trial court knows that it has an informational advantage vis a vis the appellate court in findings of fact, but does not have a considerable advantage in discretionary rulings.99 If this theory is true, then judges 98 Louis, supra note 10, at 1040 ("[The] appellate court is also likely to be more expert and reliable in matters of procedure than is a single trial judge or agency whose raison d'etre is the acquisition of substantive expertise. The number of pages in the record relevant to procedural rulings also should ordinarily be fewer than those relevant to determinations going to the merits. Hence, the time required for free review of procedural questions will not ordinarily be as great as for substantive ones."). 99 These are generalizations, of course, that do not hold in every case. Some clear error contexts involve less private information, as when the factual findings depend primarily on documentary evidence, and some involve more private information. These differences vary the level of deference that is appropriate. See, e.g., Cooper, supra note 12, at 653-54 (oral testimony, credibility issues, and longer, more complex trials may call for more deference, even within clear error review, so the standard is variable). 1 8 UTAH LAW REVIEW [NO. 1 should behave differently under the two deferential standards of review, with ideological differences most pronounced in appellate review of fact-finding decisions of the trial court. The difference between the clear error review of findings of fact and the abuse of discretion review of procedural and evidentiary decisions therefore produces a third research question: do appellate court decisions that involve deferential review but no asymmetric information or that involve asymmetric information but no deferential review show less ideological effect than those that involve both? This hypothesis involves comparing deferential standards of review that involve little private information (review under the abuse of discretion standard) with standards of review that involve both deferential review and substantial amounts of private information (review of fact-finding for clear error). C. Interaction Between Hypotheses These three research questions are not separate, discrete predictions about the judicial system, but interrelated propositions that address the role of appellate review in the context of ideology and asymmetric information. The first hypothesis tests whether deferential review increases the probability of affirmance. The second hypothesis tests whether an affirmative question to the first hypothesis induces judges to choose the standards of review based on ideological divergence between the trial and appellate courts. The third hypothesis tests whether deferential standards of review themselves are associated with ideological disagreement because of asymmetric information between the trial courts and the appellate courts. The three research questions together are necessary for a complete understanding of how deferential review works in a hierarchical court system with divergent ideological preferences. II. THE DATA Testing the hypotheses developed above required collecting data on how findings of fact and discretionary decisions are treated on appeal and the extent to which ideological divergence between trial courts and appellate courts influences that treatment. The federal judicial system provides a happy coincidence for conducting this type of empirical research on appeals. The three-judge panels in the court of appeals are assigned randomly to cases, providing a "natural experiment" for assessing the influence of ideology on various empirical questions.100 Thus, there is little risk that there will be a significant selection bias with respect to which judges within a circuit hear a particular case. To the extent ideology appears to predict the probability of reversing lower court decisions under the various standards of review, we may have greater confidence in that result than in the results of most observational studies. Investigating the three research 100 CASS R. SUNSTEIN, DAVID SCHKADE, LISA M. ELLMAN & ANDRES SAWICKI, ARE JUDGES POLITICAL? AN EMPIRICAL ANALYSIS OF THE FEDERAL JUDICIARY 4 (2006). 2012] LAW, FACT, AND DISCRETION 19 questions outlined above101 requires data on the standards of review applied in appellate cases. This analysis starts in the same place as many other analyses of appellate decision-making-the most well-known and comprehensive available database on the United States Court of Appeals developed by Donald Songer and his colleagues.102 The Songer database consists of a random sample of United States Court of Appeals cases between the years 1925 and 1996.103 This analysis incorporates all cases in the Court of Appeals database, and the update, to produce a database covering the years 1985 through 2000. These years were chosen to provide a relatively balanced mix of Democratic and Republican appointees to the bench. The data for years 1985 to 1996 were taken from the original Songer database, and the data for years 1997 through 2000 were taken from the update.104 The Court of Appeals database contains many variables that are useful to this empirical study but does not contain data on standards of review. To address this problem, original data was collected for this Article about the most common standards of review applied to federal district courts-de novo, clear error, and abuse of discretion. For each case, the data provide information on (1) whether the appellate court invoked one or both of the deferential standards of review-"clear error" or "abuse of discretion," and (2) whether the appellate court found the relevant standard for reversal was met (that is, the district court committed clear error), was not met (the district court did not commit clear error) or did not clearly state whether the relevant standard for reversal was met.105 The theory is that the "clear error" standard will serve as a proxy for judicial fact-finding and therefore asymmetric information of the trial court. The coding strategy makes use of the fact that federal appellate courts generally review trial court findings of fact under the "clearly erroneous" standard. This standard applies in civil cases because of Federal Rule of Civil Procedure 52(a),106 and in many criminal cases because of statutory enactments and case law. Thus, we can use the 101 See supra Part I.B. 102 Donald R. Songer, U.S. Appeals Courts Database: The Original U.S. Appeals Courts Database 1925-1996, JUD. RES. INITIATIVE, http://www.cas.sc.edu/poli/juri/ appct.htm (last updated Oct. 21, 2008). 103 See id. The cases are actually from a stratified sample by circuit rather than simple random samples. 104 Donald R. Songer, Update to the Appeals Courts Database (1997-2002), JUD. RES. INITIATIVE, http://www.cas.sc.edu/poli/juri/appct.htm (last updated July 13, 2011). 105 Using original data collected by the author, the coding was conducted by identifying each instance in which the terms "clear error," "clearly erred," "clearly erroneous," or similar unambiguous references to the clearly erroneous standard occurred. The paragraph was then read and coded as finding clear error or finding no clear error only if the court's conclusion was unambiguous. If an appellate court reviewed multiple findings and found clear error as to some and no clear error as to others, the case was coded as having found clear error. 106 FED. R. CIV. P. 52(a). This is the most cited single rule in the Federal Rules of Civil Procedure. CHARLES WRIGHT, THE LAW OF FEDERAL COURTS § 96, at 647 (4th ed. 1983). 2 0 UTAH LAW REVIEW [NO. 1 appellate court's invocation of the clearly erroneous standard as an indication of whether the court is reviewing fact-finding of the lower court. This strategy should capture most instances of appellate review of findings of fact, at least when properly reviewed pursuant to a deferential standard of review. This is because parties are required to identify the standard of review in their briefs,107 and courts usually explicitly address the standard of review in their opinions.108 Indeed, in at least one case the Supreme Court has criticized an appellate court for not explicitly identifying the clearly erroneous standard when it reviewed findings of fact.109 Thus, the "clearly erroneous" standard is likely a reliable indicator of whether the appellate court is reviewing the lower court's factual findings.110 In addition to whether standards of review and ideology affect appellate outcomes, one of the key questions in the theory developed in the previous section is how asymmetric information and appellate deference to trial court ideology interact. To disentangle the effects of incomplete information and deference we need a point of comparison with the clear error standard-an example of deferential appellate review where information is relatively symmetrical between the trial and appellate court. Conveniently, the "abuse of discretion" review applied to various evidentiary and procedural rulings fits this description. Abuse of discretion is a deferential standard of review, but one that takes place in the context of relatively symmetric information between the trial and appellate levels. Thus, by comparing cases reviewed under the clearly erroneous standard of review with those reviewed under the abuse of discretion standard of review, we can examine the effect of varying asymmetric information while holding deference relatively constant. A variable was added to determine if the appellate courts' applied the "abuse of discretion" standard, which is deferential like the "clearly erroneous" standard.111 Any differences between the ideological patterns in "clear error" review from "abuse of discretion" review would help to disentangle the effects of a deferential standard from those of asymmetric information. To determine the effect of ideological differences on standards of review, the analysis requires measures of judicial ideology for the district courts, the appellate panels, and the circuits. Although political scientists have attempted to trace 107 FED. R. APP. P. 28(a)(9)(B), 28(b)(5) (as amended, effective December 1, 1998). 108 MAURICE ROSENBERG, Standards of Review, in RESTRUCTURING JUSTICE: THE INNOVATIONS OF THE NINTH CIRCUIT AND THE FUTURE OF THE FEDERAL COURTS 32 (Arthur D. Hellman ed., 1990). 109 Amadeo v. Zant, 486 U.S. 214, 223-24 (1988). 110 There is, of course, always the possibility that courts selectively invoke the "clearly erroneous" standard of deferential review according to their substantive agreement or disagreement with the court below. This possibility, however, is exactly what the third hypothesis is designed to test. 111 The coding procedures applied were similar to those for the clearly erroneous standard. 2012] LAW, FACT, AND DISCRETION 21 ideological effects in judicial decision-making for many years,112 the art and science of measuring ideology for judges in the lower federal courts is in its infancy, and to say that such measures are controversial would be an understatement.113 Because this Article requires a consistent measure that can be applied to district courts and appellate courts, it follows recent work on the lower courts114 by measuring ideological preferences of judges with Giles-Hettinger- Peppers scores ("GHP scores").115 The GHP scores are ideological scores for judges that range from -1 (most liberal) to 1 (most conservative) based on ideological estimates for the political actors involved in selecting the judges-that is, the President and the Senators from the judge's state.116 The scores are often referred to as common-space NOMINATE scores because they are designed to put judges, representatives, senators, and the president on a common scale, and are based on NOMINATE scores developed by Poole and Rosenthal. 117 The scores are developed for the political actors (President and senators) based on roll-call voting in Congress,118 and then these scores are in turn used as estimates of judicial ideology. To develop a score for a particular judge, GHP uses the score for the appointing President or the Senator(s) with courtesy over the appointment, if any.119 If senatorial courtesy120 was not operative in the judge's appointment, the GHP score is the common space score of the President who appointed the judge.121 If senatorial courtesy was operative, the GHP score is the 112 See, e.g., DONALD R. SONGER, REGINALD S. SHEEHAN & SUSAN B. HAIRE, CONTINUITY AND CHANGE ON THE UNITED STATES COURTS OF APPEALS 111-13 (2000) (citing numerous studies examining predictors of judicial decisions). 113 See generally Joshua B. Fischman & David S. Law, What Is Judicial Ideology, and How Should We Measure It?, 29 WASH. U. J.L. & POL'Y 133 (2009) (identifying various problems with existing measures of judicial ideology, and developing criteria for researchers to improve the measures). 114 E.g., CROSS, supra note 63, at 19 (discussing historic method of using "the party of the president who appointed the judge as a guide to the judge's own ideology"). 115 See generally Micheal W. Giles, Virginia Hettinger & Todd Peppers, Picking Federal Judges: A Note on Policy and Partisan Selection Agendas, 54 POL. RES. Q. 623 (2001) (discussing the operation of the policy and partisan agendas in the selection of judges to the Court of Appeals). 116 Id. at 631. 117 See generally KEITH T. POOLE & HOWARD ROSENTHAL, CONGRESS: A POLITICAL-ECONOMIC HISTORY OF ROLL CALL VOTING 23-24(1997) (discussing the NOMINATE procedure for analyzing roll call data); Keith T. Poole, Recovering a Basic Space from a Set of Issue Scales, 42 AM. J. POL. SCI. 954, 977-91 (1998) (discussing Poole and Rosenthal's NOMINATE procedure). 118 POOLE & ROSENTHAL, supra note 117, at 23-24. 119 Giles, Hettinger & Peppers, supra note 115, at 631. 120 "Senatorial courtesy is an unwritten rule of the Senate that gives a senator of the same party as the president the ability to block a nominee to a district court judgeship in the senator's state." David S. Law, Appointing Federal Judges: The President, the Senate, and the Prisoner's Dilemma, 26 CARDOZO L.REV. 479 , 493-494 (2005). 121 See id. 2 2 UTAH LAW REVIEW [NO. 1 common space score of the senator with courtesy (or the mean of the two Senators if both had courtesy).122 The GHP scores are noisy but allow a continuous measure of ideological preferences that are designed to be comparable over time. In addition, the innovation of GHP over common space scores of the appointing president used in other studies123 appears to increase the accuracy of the scores,124 at least with respect to appellate judges. Although another study has suggested that for district judges presidential ideology is dominant over senatorial ideology,125 a consistent measure of preferences between the two court levels is highly desirable for conducting research on hierarchical interaction. Accordingly, the GHP scores are used throughout this Article. The measure of ideology for the district court judge is simply his or her GHP score. Following existing literature, the measure of ideology for the appellate panel is the median of the panel's GHP scores.126 The key ideological variable throughout most of the analyses is the distance between the ideology of the district court and the ideology of the appellate court (hereinafter "IdeologyDiff"), which is simply the absolute value of the difference of the two scores.127 Some summary statistics for the judges' scores in each circuit and the IdeologyDiff score are set forth in Table 1, below. Higher GHP scores represent more conservative ideological values, lower GHP scores represent more liberal values, and larger values of IdeologyDiff represent greater ideological distance between the trial court and the appellate panel. 122 Id. 123 See, e.g., Susan B. Haire, Stephanie A. Lindquist & Donald R. Songer, Appellate Court Supervision in the Federal Judiciary: A Hierarchical Perspective, 37 LAW & SOC'Y REV. 143, 156-59 (2003) (relying upon NOMINATE scores for "an estimate of presidential ideological preferences"). 124 Giles, Hettinger & Peppers, supra note 115, at 635-38. 125 Susan W. Johnson & Donald R. Songer, The Influence of Presidential Versus Home State Senatorial Preferences on the Policy Output of Judges on the United States District Courts, 36 LAW & SOC'Y REV. 657, 661-65 (2002). 126 See, e.g., CROSS, supra note 63, at 32-33 (arguing that the panel median ideology best reflects the ideology of the panel from a theoretical and empirical perspective). 127 This approach is substantially the same as that taken in a previous study of reversals using the same dataset (without standards of review). See VIRGINIA A. HETTINGER, STEFANIE A. LINDQUIST, AND WENDY L. MARTINEK, JUDGING ON A COLLEGIAL COURT 91 (2006). The authors in that study used the ideology of the opinion writer rather than of the panel for purposes of consistency with their other results, but noted that the alternative measures made little substantive difference. See id. at 135. 2012] LAW, FACT, AND DISCRETION 23 Table 1. Mean GHP Scores. District Court Appellate Panel Median IdeologyDiff First Circuit -0.072 -0.017 0.419 Second Circuit -0.117 -0.082 0.253 Third Circuit -0.052 -0.04 0.333 Fourth Circuit 0.163 0.156 0.358 Fifth Circuit 0.213 0.224 0.304 Sixth Circuit 0.116 0.076 0.412 Seventh Circuit 0.08 0.076 0.365 Eighth Circuit -0.02 0.022 0.358 Ninth Circuit 0.09 -0.063 0.402 Tenth Circuit 0.044 0.071 0.392 Eleventh Circuit 0.143 0.14 0.262 DC Circuit -0.058 0.129 0.527 A quick look at the Table shows the relative dominance of Republican appointees during the period (most of the entries are positive). As would be expected, the appellate panels are generally more moderate and less variable than the district court judges, which results from taking the median of the panel ideologies. The circuits show a fair amount of variability in their ideological tendencies, but the figures roughly parallel expectations. III. EMPIRICAL EVIDENCE ON STANDARDS OF REVIEW This Part analyzes the research questions developed above128 in light of the data described in Part II. The results presented in this Part break down into three subparts-one for each of the hypotheses presented in Part I-B. Part III-A considers the question of whether judges appear to be constrained by the standards of review or whether they appear to choose ideologically despite standards of review. Part III-B considers the question of whether judges ideologically choose among standards of review. Part III-C considers the question of whether judges choose ideologically within standards of review-in other words, does the effect of ideology play out differently between deferential review and de novo review on the one hand, or between conditions of asymmetric information and conditions of symmetric information, on the other. A. Do Standards of Review Affect Reversal Rates? The first research question was whether deferential standards of review correlate with decreased reversals and increased affirmances relative to de novo 128 See supra Part I.B. 2 4 UTAH LAW REVIEW [NO. 1 review. The legal model would suggest such a relationship, but skeptics would question the constraining force of the command of deference. Table II below presents a simple cross tabulation of the proportions of reversal and affirmance by the standard of review applied.129 Affirmed Affirmed in Part and Reversed/ Vacated in Part Reversed or Vacated Number of Cases Clearly Erroneous Only 0.616 0.165 0.220 643 Abuse of Discretion Only 0.617 0.188 0.195 1027 Both 0.583 0.274 0.142 372 Neither (De Novo) 0.584 0.104 0.313 3903 Table II does not show the clear relationship between the affirm/reverse decision expected by the legal model, but instead shows some surprisingly nuanced relationships among the standards of review and the outcome of the appeal. In particular, whether one sees standards of review as mattering depends critically on whether one is presented with the rate of "affirmance" or the rate of "reversal."130 The rate of affirmance does not vary much across the different standards of review, suggesting that deferential standards of review make little difference in appellate outcomes. The rate varies from about 58 percent in nondeferential review cases to about 62 percent in the deferential review cases, a relatively modest difference, but in the expected direction. However, in cases where both clearly erroneous and abuse of discretion standards apply, the rate of affirmance is actually lower than the cases where neither deferential standard of review applies. The situation is reversed, however, if one looks at the rate of reversal rather than the rate of affirmance, where standards of review appear to play a significant role. The reversal rate varies from about 31 percent when no deferential standard applies to 22 percent in the clearly erroneous standard cases, 20 percent in the abuse of discretion standard cases, and 14 percent when both deferential standards applied. Table II leaves the puzzling conclusion that both the rate of affirmance and the rate of reversal are minimized when both deferential standards of review apply. These basic figures make it clear that there is something between affirmance and reversal that is closely tied to deferential standards of review. 129 The analysis in this section is similar to the "outcomes analysis" of standards of review that others have performed for administrative decisions. Paul R. Verkuil, An Outcomes Analysis of Scope of Review Standards, 44 WM. & MARY L. REV. 679, 682 (2002); see also CROSS, supra note 63, at 53. 130 For this purpose, cases "vacated" are treated as reversals. 2012] LAW, FACT, AND DISCRETION 25 The missing piece in the puzzle is that not all appellate decisions "reverse" or "affirm" trial court decisions in toto; instead, a certain percentage of decisions affirm the trial court in part and reverse or vacate in part. This percentage ranges from a relatively insignificant 10 percent of the cases when no deferential standard applies to a fairly substantial 27 percent of the cases when both deferential standards apply, a significant proportionate increase. The principal effect of standards of review appears to shift cases from the "reversed" category to the "affirm in part and reverse in part" category. This means that any conclusion as to whether deferential standards of review affect the outcomes of appeals depends greatly on whether one counts decisions affirmed in part and reversed in part as "affirmances," "reversals,"131 or whether one simply omits them altogether. Indeed, omitting these "mixed" outcomes is exactly what studies of the appellate courts tend to do.132 In many cases the omission makes sense, but it seems clear that these mixed outcome decisions have an integral relationship with standards of review. As a result, any discussion of "reversal rates" or "affirmance rates" is potentially misleading without taking into account this effect. The data therefore appear to support the proposition that deferential standards of review make reversals less likely relative to other outcomes. To the extent that mixed outcomes are interpreted as affirmances rather than reversals, the results support the "legal model" that standards of review reduce the reversal rate. The conclusion does not necessarily mean that the standards of review actually constrain the ideological decision-making of judges nor does it mean necessarily that judges are not actually constrained much more than the results suggest. This is because a variety of selection effects could produce the same results in the absence of constraint, or work to attenuate the results in the presence of a strong constraint. The alternative interpretation is that deferential standards of review are more likely to be mentioned when a case is more complex and the complexity of the case significantly increases the probability of a mixed outcome.133 The affirmance rate depends not only on the appellate court's decision, but also on the strategic anticipation of that decision by trial courts and litigants.134 Civil litigants probably do not flip coins to decide whether to appeal particular issues, meaning that the cases and issues that are appealed probably depend on 131 Cross coded as "affirmed" the categories of "affirmed," "petition denied," or "appeal dismissed" as listed in the Songer database, Songer, supra note 102, and "reversed" as all other categories, Frank B. Cross, Decisionmaking in the U.S. Circuit Courts of Appeals, 91 CALIF. L. REV. 1457, 1500-01 (2003). 132 See Stefanie A. Lindquist, Wendy L. Martinek & Virginia A. Hettinger, Splitting the Difference: Modeling Appellate Court Decisions with Mixed Outcomes, 41 LAW & SOC'Y REV. 429, 430 (2007) ("[M]ost studies of appellate court outcomes routinely exclude from analysis cases where the outcome cannot be clearly classified as in favor of one party or the other or cases that have ‘ambiguous results."'). 133 Id. at 446. 134 See Schanzenbach & Fischman, supra note 63, at 408-09. 2 6 UTAH LAW REVIEW [NO. 1 self-selection based on the characteristics of the cases themselves.135 But criminal appeals provide a useful way to test the influence of selection effects. Criminal defendants in general have strong incentives to appeal, and because of public financing of criminal appeals, indigent criminal defendants in particular have no incentive not to appeal.136 As a result, criminal defendants tend to appeal using a kitchen sink approach.137 They are much more likely to appeal every possible issue no matter how small the probability of success. Thus, criminal defendants will often raise clear error on appeal even if there is no evidence of factual error, a relationship that we will use in interpreting the results below. Second, judges might strategically select standards of review, biasing the results observed in this section. If judges tended to apply deferential standards of review to decisions they agree with and nondeferential standards of review to decisions they disagree with, then deferential standards of review would show a higher affirmance rate even if they did not affect appellate review. This possibility-the second of the three research questions-is taken up in the next section. B. Do Judges Strategically Choose Among Standards of Review? The strategic theory of standards of review predicts at least two types of strategic behavior, as described in Part I. First, to the extent that appellate courts are obedient to deferential standards of review, trial courts will potentially have an incentive to use factual, evidentiary, or procedural rulings to insulate their decisions from the increased scrutiny applied under de novo review. Second, appellate courts apply deferential standards of review to decisions they agree with and nondeferential standards to decisions they do not agree with. As Frank Cross pointed out in a similar analysis of standards of review of administrative agencies: [T]he judge may have some discretion in choosing a review standard and may select the standard that best suits the judge's preferred ideological outcome. If so, the association might not demonstrate the constraining power of the review standard but only the judge's use of a deferential standard when his or her extralegal preferences are for deference."138 135 The classic article on litigant selection is George L. Priest & Benjamin Klein, The Selection of Disputes for Litigation, 13 J. LEGAL STUD. 1 (1984). This article made the well-known prediction that the success rate in litigated cases to be 50 percent because of litigant selection. Id. at 17-22. The litigant selection model has been applied to appeals by Clermont & Eisenberg, supra note 57, at 150-52. 136 See THOMAS E. BAKER, RATIONING JUSTICE ON APPEAL: THE PROBLEMS OF THE US COURTS OF APPEALS 178 (1994). 137 See Howard v. Gramley, 225 F.3d 784, 791 (7th Cir. 2000) (finding that although the "kitchen sink" approach to raising issues on appeal is widely utilized, an attorney is "clearly not incompetent" when she declines to follow the approach). 138 CROSS, supra note 63, at 53. 2012] LAW, FACT, AND DISCRETION 27 Either one of these possibilities could mean that the results above do not reflect deference, but rather strategic maneuvering by the trial or appellate courts. Testing the possibility of strategic choice among standards of review requires an examination of how judicial preferences and standards of review interact. There are two possible selection effects: (1) the trial court may choose to ground its decisions on rulings that receive deferential review, which it should do when it expects an ideologically dissimilar appellate court; and (2) the appellate court may strategically choose standards of review, preferring deferential review when the trial court is ideologically similar and de novo review when the trial court is ideologically dissimilar. The two effects would be difficult to untangle, but the random assignment of panels creates a means of distinguishing strategic trial court action from strategic appellate court action. This is because the district court does not know the identity of the appellate panel when it makes its decision, so it must make any strategic decisions based on the composition of the circuit.139 The appellate court, on the other hand, knows the identity of the trial judge when it makes its decision, so it can calculate without guesswork. Thus, to the extent trial judges manipulate standards of review, the manipulation should correlate with the district judge's distance from the ideological center of the circuit. To the extent that the appellate court manipulates standards of review, that manipulation should correlate with the district judge's distance from the ideological center of the appellate panel.140 Table III presents ideological distances for each of these standard of review categories below. Table III. Ideological Distances by Standard of Review. Average Ideological Distance from Appellate Panel Average Ideological Distance from Circuit Median141 Clearly Erroneous Only 0.362 0.332 Abuse of Discretion Only 0.367 0.327 Both 0.362 0.314 Neither (De Novo) 0.358 0.326 139 Guthrie & George, supra note 57, at 368-74 (explaining that a strategic district court would take the circuit's preferences into account in decision-making). 140 To the extent that the appellate panel anticipates and responds to the possibility of en banc reversal by the whole circuit, the two effects would be difficult to separate. There is, in fact, evidence that it is more likely that the en banc circuits discipline the panels in the rehearing when the appellate panel reverses the district court. Tracey E. George, The Dynamics and Determinants of the Decision to Grant En Banc Review, 74 WASH. L. REV. 213, 267 (1999). 141 The circuit median ideology was computed using the median ideology of the judges in each circuit in the cases in the dataset. The circuit median ideology was allowed to adjust over time, fixed within each 100 volumes of the Federal Reporter. 2 8 UTAH LAW REVIEW [NO. 1 The data do not reveal any indication that the ideological distance between the district court and the appellate court affects the probability of invoking deferential review. The relevant comparison is between the standards of review-that is, comparing within each column.142 If district judges were acting strategically, one would expect deferential review to correlate with larger distances from the circuit median, but in fact the numbers are nearly identical. If appellate panels were acting strategically, one would expect deferential review to correlate with smaller distances from the panel median, but in fact the opposite is true. The same is true even when the other variable is held constant, suggesting no support for the predicted relationship of ideological distance and the probability of applying deferential standards of review. This result is encouraging because if, indeed, this Table showed a systematic relationship between standard of review and ideology, we might worry that the relationships in the other analyses have been miss-specified or at least plagued by serious selection bias. C. Ideology within Standards of Review The results so far suggest some support for the legal model, and no support for the strategic invocation of standards of review. The main research question remains to be tested however-how standards of review and ideology interact in a judicial hierarchy. Specifically, how much do the standards of review-de novo, abuse of discretion, and clear error-explain the outcomes of appeals relative to ideology, and how does the effect of ideology differ within each standard of review? To investigate these patterns in a more systematic way this Part uses multinomial logistic regression analysis143 to determine how selected variables correlate with the outcomes of appellate cases. The dependent variable144 is the outcome of the appeal, which can take one of three values: (1) affirmed, (2) reversed, or (3) affirmed in part and reversed in part.145 The key independent 142 The average ideological distance from the district court to the circuit court median is smaller than the average distance to the appellate panels because the circuit median is not as variable as the panel median. 143 In general, regression analysis is a tool that "examines the relationship between a quantitative dependent variable Y and one or more quantitative independent variables." JOHN FOX, APPLIED REGRESSION ANALYSIS, LINEAR MODELS, AND RELATED METHODS 16 (1997). Logistic regression (also called logit) is a specialized version of regression analysis used when the dependent variable is binary. See J. SCOTT LONG, REGRESSION MODELS FOR CATEGORICAL AND LIMITED DEPENDENT VARIABLES 34-35 (1997). Multinomial logistic regression is an extension of logistic regression (also known as logit) that is used when the dependent variable has more than two nominal categories. Id. at 148-51. 144 The dependent variable (also called the "outcome" or "response" variable) is the event one is attempting to explain in a regression analysis. 145 The outcomes are treated as nominal, rather than ordinal, leading to the use of multinomial logistic regression rather than ordered probit. A priori, it would seem natural to think of these categories as "ordered," with "affirm in part and reverse in part" an 2012] LAW, FACT, AND DISCRETION 29 variables146 of interest are (1) the ideological distance between the appellate court and the district court, denoted IdeologicalDiff, (2) whether the abuse of discretion standard of review applied, (3) whether clear error applied, and (4) the interaction of IdeologicalDiff with the standard of review. Table III below sets for the results for these key variables of interest. The interpretation of a multinomial logistic model is similar to that of a regular logistic regression model, except that because there are three categories of outcomes here, there are three columns for each model. The first column under each model shows how the variables affect the probability of an outcome of "Reverse" relative to an outcome of "Affirm" for that model (with no control variables for case type). The second column under each model shows how the variables affect the probability of "Mixed" (for example, affirm in part and reverse in part) versus "Affirm" for that model. The third column for each model shows how the variables affect the probability of "Reverse" versus "Mixed" for that model. The top number in each box is the coefficient for the relevant variable that relates the variable to the outcome for that column, and the bottom number in each box is the p-value for that coefficient. Positive coefficients indicate that the relevant outcome (Reverse or Mixed) becomes more likely as that variable increases and negative coefficients indicate that the relevant outcome (Reverse or Mixed) becomes less likely as that variable increases. intermediate outcome between "Reverse" and "Affirm" thereby suggesting a natural ordering. Indeed, that is exactly how another analysis of the same data analyzed the outcomes. See HETTINGER, LINDQUIST &MARTINEK, supra note 127, at 97 ("[W]e think of affirming or reversing in full as representing opposite ends of a spectrum, with affirming in part/reversing in part constituting the middle ground between them."). The data on standards of review presented above, however, show that the "Mixed" outcome of affirm in part and reverse in part is qualitatively different from the two others, not an intermediate category between the two. 146 The independent variables (also called "predictor" variables) are the variables used to explain the dependent variable. 3 0 UTAH LAW REVIEW [NO. 1 Table III. Results from Multinomial Logit Model 1. Model 1 Reverse versus Affirm Mixed versus Affirm Reverse versus Mixed Intercept -0.605*** (<0.001) -1.576*** (<0.001) 0.971*** (<0.001) Clear Error -0.643*** (<0.001) 0.067 (0.680) -0.709*** (<0.001) Abuse of Discretion -0.276* (0.046) 0.751*** (<.001) -1.027*** (<0.001) IdeologyDiff 0.005 (0.970) -0.278 (0.167) 0.283 (0.188) IdeologyDiff x Clear Error 0.775* (0.020) 0.989** (0.005) -0.215 (0.607) IdeologyDiff x Abuse of Discretion -0.723* (0.019) -0.768* (0.019) 0.045 (0.910) P-values in parentheses. *** Denotes p < .001 ** Denotes p < .01 * Denotes p < .05 Table III shows strong support for some aspects of the legal model of standards of review, but echoes the initial conclusion in Subpart A that the relationships are more complex than the legal model suggests. On the one hand, both deferential standards of review clearly correlate with a decreased probability of an outcome of Reverse relative to Affirm or Mixed, a finding consistent with the legal model. But the standards of review act differently from one another in their effect on the Mixed outcome. The abuse of discretion standard of review is strongly associated with a much higher probability of a Mixed outcome relative to either Affirm or Reverse. The clear error standard, in contrast, is strongly associated with a higher probability of Mixed relative to Reverse, but only with a slightly higher probability of Mixed relative to Affirm. Thus, the clear error standard appears to greatly reduce the probability of Reverse, shifting the likely outcome toward both Affirm and Mixed. The abuse of discretion standard appears to greatly increase the probability of Mixed, shifting the likely outcome not only away from Reverse, but also away from Affirm. Thus, the two deferential standards of review appear to behave differently, suggesting that either the context or the command of deference or both are different between the two. The next set of variables incorporates the role of ideology by considering how IdeologyDiff relates to the other variables. Recall that IdeologyDiff is the estimated degree of ideological disagreement between the trial and appellate courts in the case. The IdeologyDiff variable standing by itself has very little and unclear impact in Model 1. This means that in cases where no deferential standard of review applies (de novo), as the ideological distance between the trial court and the 2012] LAW, FACT, AND DISCRETION 31 appellate court increases, the probability of the appellate court reversing, vacating, or reversing in part does not change much. In a sense, this finding is surprising given the emphasis of ideology in judicial politics literature, and indeed some studies find that ideology is a strong predictor of appellate court outcomes.147 At the same time, however, political science scholars have largely failed to find the large ideological effects that are present in the Supreme Court at the appellate court level.148 Indeed, a recent study using this same data as this Article found no ideological influence over the reverse/affirm decision, a result that ran contrary to what the study's authors had predicted.149 The prior study did not, however, look at ideology in the context of deferential standards of review, and that is where Table III suggests a principal source of the ideological conflict lies. When clear error review applies, differences in ideology between the trial court and the appellate courts makes Reverse and Mixed more likely relative to Affirm. When abuse of discretion applies, differences in ideology between the trial court and the appellate court actually make Affirm more likely relative to Reverse or Mixed. Thus, the ideological difference between the trial court and appellate court only appears to reduce the probability of an Affirm decision when much more when the deferential clear error standard applies than when the de novo standard applies, a highly counterintuitive conclusion. To dig deeper into this surprising result, Model 2 breaks down the effect of ideology by case type, adding a set of control variables for three broad case types in the database: (1) criminal, (2) civil rights and civil liberties,150 and (3) other.151 147 See, e.g., Richard Revesz, Environmental Regulation, Ideology, and the D.C. Circuit, 83 VA. L. REV. 1717, 1717-21, 1735 (1997) (finding ideological voting by party of appellate judges in environmental cases in the District of Columbia Circuit); Cross, supra note 131, at 1462-90, 1515 (finding support for both a "political model" and a "legal model" of decision-making on the Court of Appeals, with the "legal model" providing more explanatory power). 148 Frank Cross has hypothesized that the difference results in part from the fact that the Supreme Court mostly reviews "‘close cases,' in which the law is relatively evenly balanced on each side of a dispute." CROSS, supra note 63, at 17. He explains that "[t]he circuit courts, by contrast, hear many more cases and cannot pick and choose the cases they decide," meaning the "much broader set of cases is unlikely to be so evenly balanced on the law, so circuit court judges may not be forced to consider ideology, as Supreme Court justices are." Id. 149 See HETTINGER, LINDQUIST &MARTINEK, supra note 127, at 98. 150 The Songer database codes for civil rights cases but does not have a separate single variable for civil liberties cases. See Donald R. Songer, The United States Courts of Appeals Data Base Documentation for Phase 1, JUD. RES. INITIATIVE, 77-78, http://www.cas.sc.edu/poli/juri/cta96_codebook.pdf (last visited Apr. 14, 2012). Civil liberties cases were coded as those falling into the categories of First Amendment, Due Process, and Privacy. See id. at 77-89. Note that civil rights is the dominant type of case in this category, accounting for 935 cases versus 295 for all civil liberties cases combined. 3 2 UTAH LAW REVIEW [NO. 1 There are two reasons for controlling for these specific areas of law. First, there is the potential that the area of law, rather than the clear error standard, is driving the ideological effects found in Table III. If different areas of law apply the standards of review with different frequencies, and different areas of law have different levels of ideological disagreement, then ideology may appear to be activated by the standard of review when really the different areas of law drive the result. Indeed, existing work on ideology in the appellate courts tends to rely rather extensively on using only civil liberties cases, something that has been criticized by legal scholars.152 The reason is probably that civil rights/civil liberties cases tend to show stronger ideological effects than most other categories with respect to the case outcome, both in appellate courts and in district courts.153 So it is necessary to control for civil rights and civil liberties cases to ensure that the effects observed in Table III do not disappear. The reasons for controlling for criminal cases are probably even more compelling than for civil rights and liberties cases. As with civil rights and civil liberties cases, criminal cases may tend to activate ideological disagreement more than the average noncriminal case.154 But unlike civil rights and civil liberties cases, criminal cases tend to be affirmed much more frequently than noncriminal cases,155 raising the possibility of bias in the effect of standards of review. There are a variety of reasons for each of these facts, but one of the most important ones in the current context is that the incentives for appeal are quite different in criminal cases.156 Criminal defendants are much more likely to appeal than defendants in other types of cases, and are more likely to appeal than the prosecution in criminal cases themselves.157 These unique features of criminal cases have the potential to bias the results if not controlled for in the model. Therefore, Model 2 includes variables for whether the case was a criminal case, a civil rights/civil liberties case, and also includes interaction effects of each 151 The "other" category includes the remaining three case types in the database, (1) labor relations, (2) economic activity and regulation, and (3) miscellaneous. See id. at 85- 87. The second category amounts for the vast majority of cases in the "other" group. Id. 152 Barry Friedman, Taking Law Seriously, 4 PERSP. ON POL. 261, 271 (2006) (noting the overreliance on civil liberties cases in political science scholarship). 153 C.K. ROWLAND & ROBERT A. CARP, POLITICS AND JUDGMENT IN FEDERAL DISTRICT COURTS 37 (1996) (explaining that in the district courts, "partisan effects are concentrated primarily in the realm of civil rights and liberties, and that partisan polarization in this realm is increasing"). 154 See id. at 39 (noting that district court judges "seem[] to have split along ‘political' lines more often on criminal justice and on civil rights matters than they did with other types of cases"). 155 See, e.g., Guthrie & George, supra note 57, at 362-63 (comparing the affirmance rate of criminal cases to cases overall). 156 See supra notes 134-137 and accompanying text. 157 See John Scalia, Bureau of Justice Statistics, Special Report, Federal Justice Statistics Program, Federal Criminal Appeals, 1999 with Trends 1985-99, at 1 (1999), available at http://bjs.ojp.usdoj.gov/content/pub/pdf/fca99.pdf (reporting that in federal criminal appeals "95% of appeals were filed by the defendant; 5% by the government"). 2012] LAW, FACT, AND DISCRETION 33 of these case types with IdeologyDiff. This interaction variable allows us to check whether the ideological effects of standards of review found above are merely the result of different standards of review in different areas of law. These control variables are not extensive, but there are reasons to believe that extensive control variables are not necessary,158 and additional control variables were tested and did not change the results presented in Table III or IV in any significant respect.159 158 The main reason is that random assignment of appellate judges to panels tends to mitigate many of the selection effects that we would otherwise want to control for. Thus, even with relatively few controls, any ideological effects seem plausible as causal variables. 159 The principal alternative model included control variables for the circuit in which the appeal was heard. One reason for controlling by circuit is that different circuits have different case law about when to apply the various standards, as well as about what level of deference the standards themselves require. Second, the circuits may differ systematically in terms of the types of cases that reach the courts of appeals, whether because of social factors, see SONGER, supra note 112, at 49-50, or institutional factors, see id. at 53-54. Indeed, differences in ideology among circuits continue even when regional effects are controlled. See id. at 125-28. Third, the circuits vary ideologically from one another, and there are varying mixes of ideological actors within circuits. The results when the circuits were controlled were qualitatively the same as those presented. 3 4 UTAH LAW REVIEW [NO. 1 Table IV. Results from Multinomial Logit Model 2. Model 2 Reverse versus Affirm Mixed versus Affirm Reverse versus Mixed Intercept -0.264** (0.002) -1.369*** (<0.001) 1.105*** (<0.001) Clear Error -0.482** (0.003) 0.269 (0.108) -0.752*** (<0.001) Abuse of Discretion -0.251 (0.073) 0.786*** (<0.001) -1.037*** (<0.001) IdeologyDiff -0.452* (0.015) -0.555* (0.036) 0.104 (0.716) IdeologyDiff x Clear Error 0.739* (0.031) 0.867* (0.017) -0.128 (0.766) IdeologyDiff x Abuse of Discretion -0.720* (0.020) -0.803* (0.015) 0.083 (0.835) CivilRightsAndLiberties -0.380* (0.008) 0.156 (0.340) -0.536** (0.004) CivilRightsAndLiberties x IdeologyDiff 0.994*** (<0.001) 0.341 (0.379) 0.653 (0.120) Criminal -0.816*** (<0.001) -0.844*** (<0.001) 0.028 (0.878) Criminal x IdeologyDiff 0.653* (0.019) 0.694 (0.057) -0.041 (0.921) P values in parentheses. *** Denotes p < .001 ** Denotes p < .01 * Denotes p < .05 The results in Table IV show that the findings from the original are robust against controls for the area of law. Only one coefficient (IdeologyDiff for Reverse versus Affirm) changed its sign between the two models, and the reason can be discerned by looking at Table IV. Table IV disaggregates the effects of ideology by case type and shows that ideology has different effects in different areas of law. The criminal and civil rights/civil liberties cases show a fairly strong ideological effect in the expected direction. In the criminal category, ideological distance between the trial and appellate courts makes Reverse or Mixed much more likely relative to Affirm. In the civil rights/civil liberties category, ideological distance between the trial and appellate courts makes Reverse much more likely relative to Affirm, but has less of an effect on Mixed versus Affirm. The ideological nature of civil rights decisions in particular is one that the literature has repeatedly 2012] LAW, FACT, AND DISCRETION 35 demonstrated.160 The existing literature on criminal cases does not show as strong an ideological effect,161 but the results here clearly show not only that there is such an ideological effect, but also that this effect is distinct from the effect of clear error. In contrast, in the "other" cases, ideological distance between the trial and appellate courts is actually associated with a statistically significant increase in the probability of Affirm relative to Reverse and Mixed, counter to what one might expect. Thus, in the more controversial contexts of criminal law and civil rights/civil liberties, ideological distance between the trial and appellate courts produces the expected result-lower probability of Affirm and higher probability of reversing or vacating, at least in part. In contrast, in the "other" cases- primarily economic issues and diversity cases-ideological distance between the trial and appellate courts is actually associated with a higher probability of Affirm. This is why the overall IdeologyDiff effect in Model 1 was close to zero, because the ideological and nonideological case types balance each other out.162 The results so far show that the clear error standard of review is associated with more ideological behavior between the trial and appellate courts than abuse of discretion or de novo review. Moreover, given the random assignment to panels, the interpretation of ideology as having a causal effect seems quite plausible. There are not many persuasive alternative explanations for why a president or senator's common-space ideology score would predict reversals of findings of fact.163 The results do not necessarily indicate, however, that ideology predicts when the appellate panel will actually reverse the findings of fact themselves. Instead, the results indicate simply that the appellate panel is more likely to reverse the judgment in such cases. It could be, for example, that the appellate panel simply "finds" some other way to reverse the judgment when it dislikes the findings of fact. The context of clear error review may create opportunities for ideological disagreement to translate into reversals on other grounds. Similarly, the abuse of discretion can attenuate ideological disagreement on the reverse-affirm decision without actually implying that cases are affirmed on that ground. The conclusions 160 For example, in a prominent recent study by Cass Sunstein and his colleagues all civil rights related categories showed an ideological effect of the party of the appointing president. See ELLMAN, SAWICKI, SCHKADE & SUNSTEIN, supra note 100, at 17-46, 54-57. 161 See, e.g., id. at 47--54 (showing no significant ideological effect in criminal appeals). 162 Although the other study looking at the same data controlled for civil rights/civil liberties claims, see HETTINGER, LINDQUIST &MARTINEK, supra note 127, at 94. The study did not examine the interaction of civil rights/civil liberties cases with ideology, which may explain why they did not find this ideological effect. 163 It is conceivable that within a circuit there is some regional effect that bleeds into both the common-space scores and the propensity to reverse findings of fact. This interpretation, however, would require some explanation why regional animosity would not affect the reverse/affirm decision or the abuse of discretion decision in similar ways, which it does not. 3 6 UTAH LAW REVIEW [NO. 1 so far have been based on coding appellate outcomes and standards of review without actually engaging with the court's decisions on those standards of review. The information in Table V and VI, therefore, provides corroborating evidence for the relationships described above based on the additional original data developed as part of this project. As described in Part II above, each case invoking clear error review was read to determine whether the appellate decision unambiguously stated the lower court fact-finding was clearly erroneous, was not clearly erroneous, or did not unambiguously state one way or the other.164 Each case invoking abuse of discretion review was read to determine whether the court unambiguously stated the lower court abused its discretion, unambiguously stated the lower court did not abuse its discretion, or did not unambiguously state one way or the other.165 Table V presents the relationship of ideology to clear error review of findings of fact and Table VI presents the relationship of ideology to abuse of discretion review. 164 See supra note 105 and accompanying text. 165 Recall that the cases were coded as ruling a finding was "clearly erroneous" only if the court explicitly said the finding was "clearly erroneous" or another similarly unambiguous statement. The same applied to whether the finding was "not clearly erroneous" and whether or not a ruling was an "abuse of discretion." See supra notes 105- 111 and accompanying text. If the same case found one ruling to be clear error or abuse of discretion and found another ruling to be not clear error or not an abuse of discretion,