(1319)
_________________
ARTICLE
_________________
DELIBERATION AND STRATEGY ON THE UNITED STATES
COURTS OF APPEALS: AN EMPIRICAL EXPLORATION
OF PANEL EFFECTS
PAULINE T. KIM
†
Recent studies have established that the decisions of a federal court of appeals
judge are influenced not only by the preferences of the judge, but also by
the preferences of her panel colleagues. Although the existence of these “panel
effects” is well documented, the reasons that they occur are less well understood.
Scholars have proposed a number of competing theories to explain panel effects,
but none has been established empirically. In this Article, I report an empirical
test of two competing explanations of panel effects—one emphasizing deliberation
internal to a circuit panel, the other hypothesizing strategic behavior on the
part of circuit judges. The latter explanation posits that court of appeals
judges act strategically in light of the expected actions of others and that, therefore,
panel effects should depend upon how the preferences of the Supreme Court
or the circuit en banc are aligned relative to those of the panel members. Analyzing
votes in Title VII sex discrimination cases, I find no support for the theory
that panel effects are caused by strategic behavior aimed at inducing or
avoiding Supreme Court review. On the other hand, the findings strongly suggest
that panel effects are influenced by circuit preferences. Both minority and
†
Associate Dean for Research and Faculty Development and Professor of Law,
Washington University School of Law, St. Louis. I am grateful to Cass Sunstein for
making his data available, to Christina Boyd for sharing the expanded data set, to Ryan
Black for assisting with the statistical analysis, and to Jessica Smith for able research assistance.
Thanks also to Andrew Martin, Margo Schlanger, David Law, Bryan Lammon,
Katherine Barnes, Scott Baker, Josh Fischman, Aaron Bruhl, and participants in workshops
at Washington University Law School, Emory Law School, University of Alabama
Law School, University of Kansas Law School, and in the 2008 Conference on Empirical
Legal Studies at Cornell Law School for helpful comments on this paper.
1320 University of Pennsylvania Law Review [Vol. 157: 1319
majority judges on ideologically mixed panels differ in their voting behavior depending
upon how the preferences of the circuit as a whole are aligned relative
to the panel members. This study provides evidence that panel effects do not
result from a dynamic wholly internal to the three judges hearing a case, but
are influenced by the environment in the circuit as a whole as well.
INTRODUCTION....................................................................................1320
I. COMPETING EXPLANATIONS ........................................................1329
A. Panel Effects .......................................................................1329
B. Theoretical Accounts ...........................................................1331
II. TESTING THE STRATEGIC ACCOUNT ............................................1339
A. Existing Evidence................................................................1339
B. Constructing an Empirical Test ...........................................1342
C. Data and Empirical Analysis...............................................1350
III. RESULTS.......................................................................................1358
A. The Supreme Court as Reviewing Court................................1358
B. The Circuit En Banc as Reviewing Court .............................1361
IV. DISCUSSION AND IMPLICATIONS...................................................1367
A. The Role of the Supreme Court and
the Circuit En Banc ............................................................1367
B. Internal Panel Dynamics.....................................................1370
CONCLUSION........................................................................................1374
APPENDIX.............................................................................................1376
INTRODUCTION
One of the central institutional features of the United States
courts of appeals is the use of judicial panels to decide cases. Rather
than having a single appellate judge decide each appeal, or even having
a group of appellate judges deciding in isolation and tallying their
votes, the appeals process is specifically structured to promote a collaborative
form of decision making.
1
Three appellate judges are assigned
to decide a case together, and they typically share their background
research, sit together as a panel to hear oral arguments, meet
to discuss their views, and issue a single opinion resolving the appeal.
2
Of course, not all cases are typical, and judges sometimes dissent or
1
For a detailed description of the organizational structure of the federal appellate
courts, see JONATHAN MATTHEW COHEN, INSIDE APPELLATE COURTS: THE IMPACT OF
COURT ORGANIZATION ON JUDICIAL DECISION MAKING IN THE UNITED STATES COURTS
OF APPEALS ch. 2 (2002).
2
See id. ch. 5 (discussing the interaction among the judges within a court of ap-
peals).
2009] Empirical Exploration of Panel Effects 1321
concur separately. These occurrences are relatively infrequent, however,
and cases involving separate opinions are viewed as deviations
from the usual model of appellate decision making. Thus, as D.C.
Circuit Judge Harry Edwards put it, “judging on the appellate bench is
a group process.”
3
As a matter of institutional design, why are federal appellate
courts structured in this way? Certainly it is not for the sake of efficiency,
as the same number of judges sitting alone could decide appeals
more quickly than when sitting with two of their colleagues.
Most explanations focus on the quality of decision making.
4
Kornhauser
and Sager, for example, assert that increasing the number of
judges making a decision will increase the probability that a court will
reach a correct decision.
5
So long as each judge is more likely than
not to decide correctly, a correct outcome is more likely whenever a
group of judges decides by majority vote.
6
Others have suggested that
this error-reducing effect results from the exchange of ideas and information
that occurs during the process of deliberation.
7
For example,
Judge Edwards describes the interactions among judges on an appellate
panel as “a process of dialogue, persuasion, and revision”
8
that
enables them to “find common ground and reach better decisions.”
9
From an empirical perspective, it is difficult to test these claims in
the absence of consensus regarding what makes one decision “better”
than another. However, scholars have collected considerable evidence
suggesting that decision making by a federal court of appeals judge
sitting on a three-judge panel differs from what one might expect
3
Harry T. Edwards, The Effects of Collegiality on Judicial Decision Making, 151 U. PA.
L. REV. 1639, 1656 (2003).
4
See, e.g., Evan H. Caminker, Sincere and Strategic Voting Norms on Multimember
Courts, 97 MICH. L. REV. 2297, 2362 (1999) (noting the argument that the purpose of a
multimember appellate court is to increase accuracy); Edwards, supra note 3, at 1640-
41 (arguing that collegiality on appellate panels enables judges to “reach better decisions”);
Douglas H. Ginsburg & Donald Falk, The Court En Banc: 1981–1990, 59 GEO.
WASH. L. REV. 1008, 1035-36 (1991) (suggesting that decisions involving more judges
are more likely to be correct); cf. Michael E. Solimine, Ideology and En Banc Review, 67
N.C. L. REV. 29, 40 (1988) (reporting that en banc review is justified by the belief that
the involvement “of more judges leads to sounder decisions”).
5
Lewis A. Kornhauser & Lawrence G. Sager, Unpacking the Court, 96 YALE L.J. 82,
98 (1986).
6
Id. at 97-99.
7
See, e.g., Caminker, supra note 4, at 2372 (suggesting that collegial deliberation
will enhance the accuracy of decision making).
8
Edwards, supra note 3, at 1661.
9
Id. at 1641.
1324 University of Pennsylvania Law Review [Vol. 157: 1319
Although the existence of panel effects is well documented, the
reasons that they occur are not clearly understood. Scholars have
proposed a number of explanations, but none of these theories has
been conclusively established.
18
This Article empirically explores when
panel effects occur in an effort to better understand why they occur.
More specifically, it offers an empirical test of two competing types of
explanations: deliberative and strategic.
16
In nearly all of these issue areas, Sunstein et al. found evidence of both ideological
voting and panel effects. The exceptions to this general pattern were cases involving
criminal appeals, takings of private property, punitive damage awards, standing
to sue, and Commerce Clause challenges. In these areas, they found no difference in
the voting patterns of judges based on party affiliation. SUNSTEIN ET AL., ARE JUDGES
POLITICAL?, supra note 10, at 48. In cases involving abortion restrictions and capital
punishment, however, they found that although judges vote ideologically, their votes
do not appear to be influenced by their colleagues. Cases involving gay and lesbian
rights seemed to exhibit a similar pattern of ideological voting, but no influence from
panel composition; these cases, however, are too few in number to draw any firm conclusions
about whether panel effects are present or not. Id.
17
Republican appointees vote to uphold affirmative action programs 37% of the
time when sitting on all-Republican-appointee panels, 49% of the time when sitting
with one Republican appointee and one Democratic appointee, and 65% of the time
when sitting with two Democratic appointees. Sunstein et al., Ideological Voting, supra
note 10, at 319. For Democratic appointees, the reverse pattern holds: 82% vote in
favor of upholding affirmative action programs on an all-Democrat-appointee panel,
80% when sitting with one Democratic appointee and one Republican appointee, and
61% when sitting with two Republican appointees. Id.
18
See, e.g., CROSS, supra note 12, at 152 (explaining that the results of the earlier
studies could not explain why panel effects occur); Revesz, supra note 10, at 1755-56
(stating that his analysis cannot conclusively disentangle competing hypotheses); Sunstein
et al., Ideological Voting, supra note 10, at 307 (explaining that the data are consistent
with several different hypotheses for the causes of panel effects).
from that judge sitting alone. This phenomenon—commonly
referred to as “panel effects”—has been documented in a wide variety
of legal contexts.
***
2009] Empirical Exploration of Panel Effects 1325
By deliberative explanations, I mean to identify those theories that
emphasize the internal exchanges that occur among panel members
and the potential for these exchanges to influence a judge’s vote. For
purposes of the empirical test undertaken here, the exact mechanism
by which judges influence one another is not critical. It may be the
case that they come to persuade one another through the exchange of
information and the power of reasoned argument.
19
Alternatively,
psychological mechanisms—such as conformity pressures or group polarization—may
be operative, leading judges to change their minds
when confronted with the opinions of their colleagues.
20
The critical
point, for purposes of this study, is that pure deliberative accounts attribute
panel effects to the dynamics internal to the members of a
panel, rather than to any interaction with other actors in the judicial
system.
By contrast, strategic theories explain observed panel effects as the
result of strategic behavior by appellate judges.
21
These theories posit
that when deciding cases, individual judges advance their goals not
simply by exercising their discretion in a manner consistent with their
policy preferences, but by taking into account the likely responses of
other actors as well.
22
Rather than naïvely voting their preferences,
19
See, e.g., Edwards, supra note 3, at 1600 (arguing that a judge’s initial views may
shift through the process of collegial deliberation); Farhang & Wawro, supra note 12,
at 308 and sources cited therein (describing the deliberative model of panel decision
making).
20
See, e.g., SUNSTEIN ET AL., ARE JUDGES POLITICAL?, supra note 10, at 63-78.
21
See, e.g., VIRGINIA A. HETTINGER, STEFANIE A. LINDQUIST & WENDY L. MARTINEK,
JUDGING ON A COLLEGIAL COURT 61 (2006) [hereinafter HETTINGER ET AL., COLLEGIAL
COURT]; Cross & Tiller, supra note 10, at 2159.
22
In recent years, strategic theories of judicial behavior have become prominent
in the political science and legal literatures. See generally CROSS, supra note 12, at 94-
122 (testing various strategic theories of court of appeals decision making); Gregory A.
Caldeira, John R. Wright & Christopher J.W. Zorn, Sophisticated Voting and Gate-Keeping
in the Supreme Court, 15 J.L. ECON. & ORG. 549 (1999) (testing empirically whether Supreme
Court Justices engage in strategic voting in certiorari decisions); William N.
Eskridge, Jr., Overriding Supreme Court Statutory Interpretation Decisions, 101 YALE L.J. 331
(1991) (describing strategic interactions between the Supreme Court, Congress, and
the President in which each tries to impose its policy preferences in light of the expected
responses of other players); Forrest Maltzman & Paul J. Wahlbeck, Strategic Policy
Considerations and Voting Fluidity on the Burger Court, 90 AM. POL. SCI. REV. 581 (1996)
(testing empirically whether Supreme Court Justices act strategically in changing their
votes between the initial conference and final vote); Matthew McCubbins, Roger Noll
& Barry Weingast (McNollgast), Politics and the Courts: A Positive Theory of Judicial Doctrine
and the Rule of Law, 68 S. CAL. L. REV. 1631 (1995) (modeling judicial decision
making as a product of strategic interactions between upper and lower courts); Donald
R. Songer, Jeffrey A. Segal & Charles M. Cameron, The Hierarchy of Justice: Testing a
1326 University of Pennsylvania Law Review [Vol. 157: 1319
court of appeals judges are hypothesized to act with an eye to the expected
behavior of the Supreme Court, the circuit sitting en banc, and
their panel colleagues.
23
An appellate judge will decide whether to
vote her sincere preference or to accommodate the views of her colleagues
based on her beliefs about the likelihood of further review
and the probable outcome if the case is reviewed. Unlike purely deliberative
explanations, strategic theories suggest that panel effects will
depend upon the preferences of the Supreme Court and/or the circuit
as a whole, and not just upon the preferences of the three judges
comprising an appellate panel.
Strategic theories play an important role in some accounts of the
federal judicial hierarchy. Many scholars have suggested that the risk
of reversal assures that lower court judges follow the doctrines set out
in Supreme Court precedent, even those with which they disagree.
24
However, given the tens of thousands of cases decided by the courts of
appeals each year, the Supreme Court’s limited reversal power can
only be effective if it has some mechanism for identifying appropriate
cases for review. One hypothesis is that court of appeals judges dissent
in order to signal to the Supreme Court that certain cases deviate
from established doctrine and should be reviewed.
25
Other scholars
have described the relationship between a circuit court and a threejudge
panel in a similar manner.
26
Just as the Supreme Court moniPrincipal-Agent
Model of Supreme Court–Circuit Court Interactions, 38 AM. J. POL. SCI. 673
(1994) (modeling the interaction between the Supreme Court and lower federal
courts as a principal-agent relationship); Paul J. Wahlbeck, James F. Spriggs, II &
Forrest Maltzman, Marshalling the Court: Bargaining and Accommodation on the United
States Supreme Court, 42 AM. J. POL. SCI. 294 (1998) (testing empirically whether Supreme
Court opinions are written strategically based on an examination of draft opinions).
23
Cross & Tiller, supra note 10, at 2156; see also HETTINGER ET AL., COLLEGIAL
COURT, supra note 21, at 40-41.
24
See, e.g., Charles M. Cameron, Jeffrey A. Segal & Donald Songer, Strategic Auditing
in a Political Hierarchy: An Informational Model of the Supreme Court’s Certiorari Decisions,
94 AM. POL. SCI. REV. 101, 102 (2000) (arguing that lower court judges care
about reversal because they care about the disposition of cases and their professional
reputations); Tracey E. George & Albert H. Yoon, The Federal Court System: A PrincipalAgent
Perspective, 47 ST. LOUIS U. L.J. 819, 822 (2003) (describing reversal as the Supreme
Court’s “obvious mechanism of control over lower court judges”); Songer, Segal
& Cameron, supra note 22, at 680 (theorizing that lower court judges will “shirk” less
when the likelihood of reversal is high).
25
See Cross & Tiller, supra note 10, at 2173; see also HETTINGER ET AL., COLLEGIAL
COURT supra note 21, at 76-77.
26
See, e.g., Tracey E. George, The Dynamics and Determinants of the Decision To Grant
En Banc Review, 74 WASH. L. REV. 213, 245 (1999) (noting “an implicit but incomplete
agency agreement” between the circuit court and panel).
2009] Empirical Exploration of Panel Effects 1327
tors and occasionally reverses the decisions of the lower federal courts,
a circuit sitting en banc can review and revise a panel decision that is
inconsistent with circuit precedent or norms. This form of monitoring
is costly, however, and so scholars have suggested that the circuit
will rely on signals, such as the presence of a dissenting opinion, to
identify which panel decisions warrant closer scrutiny.
27
In order to test these two competing explanations for panel effects,
I begin with the observation that strategic accounts—unlike
purely deliberative ones—predict that appellate voting behavior will
be influenced by interactions with a reviewing court. More specifically,
if appellate judges act strategically—with an eye to the likely response
of the Supreme Court or the circuit en banc—then observed
panel effects should differ depending upon how the preferences of
the appellate judges on the panel are aligned relative to those of the
Supreme Court or the circuit as a whole. By contrast, if purely deliberative
explanations are true, the preferences of the Supreme Court
or the circuit as a whole should have no systematic impact on whether
or when panel effects are observed.
In the empirical test described here, I analyze data about judges’
votes in Title VII sex discrimination cases decided by the U.S. courts
of appeals.
28
Sex discrimination cases are often perceived to be ideologically
contested, and scholars have documented the existence of
both ideological voting and panel effects in these types of cases.
29
Most prior studies of panel effects have used the party of the appointing
President as a proxy for judicial ideology
30
and then have compared
the voting records of Republican-appointed and Democratappointed
judges across different panel compositions. In this study, I
follow the convention of using the party of the appointing President
to define ideological alignments—for example, I assume that a Re-
27
See HETTINGER ET AL., COLLEGIAL COURT, supra note 21, at 76-77; George, supra
note 26, at 247.
28
See Section II.C., infra, for a more detailed description of the data.
29
See, e.g., SUNSTEIN ET AL., ARE JUDGES POLITICAL?, supra note 10, at 30-31 (finding
evidence of both ideological voting and panel effects in sex discrimination cases);
Peresie, supra note 12, at 1768-69 (finding that both judge gender and judicial ideology
significantly affect outcomes in Title VII sex discrimination and harassment cases);
Boyd, Epstein & Martin, supra note 12, at 20 (documenting large panel effects in sex
discrimination cases).
30
E.g., SUNSTEIN ET AL., ARE JUDGES POLITICAL?, supra note 10; Cox & Miles, supra
note 12, at 3; Cross & Tiller, supra note 10, at 2168; Miles & Sunstein, supra note 12, at
830; Revesz, supra note 10, at 1718-19; Sunstein et al., Ideological Voting, supra note 10, at
302 n.1.
1328 University of Pennsylvania Law Review [Vol. 157: 1319
publican-appointed judge sitting with two Democrat-appointed judges
is in the “ideological minority,” while the two Democratic appointees
are the “majority” judges on that panel.
Unlike prior studies, however, mine does not rely on the “percent
liberal” vote to measure judges’ voting behavior. Instead, I examine
the extent to which judges vote counter-ideologically—that is, in a direction
opposite to what a naïve ideological model would predict.
This methodological innovation permits a focus on the central phenomenon
of interest: the changing likelihood that a judge will vote
counter to a naïve ideological prediction depending upon the panel
composition. In the empirical test, I examine whether observed panel
effects—the changes in the likelihood of a counter-ideological vote
under different voting conditions—are contingent upon the preferences
of the Supreme Court or the circuit en banc.
Using this method, I find no evidence that panel effects are influenced
by the relative preferences of the Supreme Court. More specifically,
I observe no difference between the voting patterns of either
minority or majority judges on mixed panels regardless of whether the
minority judge is more closely aligned with the Supreme Court or with
the panel majority. This finding casts doubt on one explanation of
hierarchical control—namely, the theory that appellate judges’ voting
behavior is motivated by the desire to signal noncompliant decisions
to the Supreme Court. On the other hand, I find evidence that the
tendency of appeals court judges to be influenced by their panel colleagues
does depend on how the preferences of the circuit court as a
whole are aligned relative to those of the panel members. When a
minority judge on a panel is ideologically closer to the circuit as a
whole than to the panel majority, the majority judges are less likely to
vote in a stereotypically ideological direction, while the minority judge
is more likely to do so. This result is consistent with a strategic explanation
for panel effects, although the exact mechanism by which circuit
preferences influence panel behavior remains uncertain. What
the results do indicate is that panel effects are not the result of a dynamic
wholly internal to the three-judge panel, but are influenced by
the circuit environment.
This Article proceeds as follows: Part I surveys the competing
theoretical explanations that have been offered to explain panel effects.
Part II explains the limitations of existing empirical tests and
then describes my approach for testing strategic accounts of panel decision
making. In Parts III and IV, I present the results of the empirical
tests and then consider the implications of my findings.
2009] Empirical Exploration of Panel Effects 1329
I. COMPETING EXPLANATIONS
A. Panel Effects
As described more fully in Part II, this study analyzes data on
judges’ votes in Title VII sex discrimination cases. In the analysis and
discussion that follow, I characterize a vote in favor of the sex discrimination
plaintiff as “liberal” and a vote against the plaintiff as
“conservative.”
31
Table 1 shows that, as one might expect, the percentage
of cases with a liberal outcome varies depending upon the
composition of the panel.
Table 1: Federal Court of Appeals Decisions in Sex Discrimination
Cases, 1995–2002, by Panel Composition
Panel
Composition
Number of
Observations
Percent Liberal
Outcomes
RRR 186 25.8%
RRD 354 38.4%
RDD 199 49.2%
DDD 48 79.2%
Table 2 further breaks down the data. Consistent with prior studies,
it shows that Democratic appointees vote in favor of plaintiffs in
these cases more often than Republican appointees (51.9% of the
time as compared with 34.2% of the time), but that judges’ votes are
influenced by the partisan affiliation of the other members of the
31
This treatment is consistent with prior studies of judicial decision making in sex
discrimination and Title VII cases. See, e.g., SUNSTEIN ET AL., ARE JUDGES POLITICAL?,
supra note 10, at 19 (describing a vote for a plaintiff in a sex discrimination case as the
“stereotypically liberal” vote); Sunstein et al., Ideological Voting, supra note 10, at 314
tbl.1 (identifying a vote for the plaintiff in sex discrimination cases as voting for the
liberal position); Boyd, Epstein & Martin, supra note 12, at 19 (treating pro-plaintiff
votes in sex discrimination cases as liberal cases).
1330 University of Pennsylvania Law Review [Vol. 157: 1319
panel as well as their own. For example, a Republican appointee sitting
with two Democratic appointees casts a liberal vote 44.2% of the
time. However, her voting pattern becomes steadily more conservative
when she sits with one other Republican appointee (37.7% liberal
votes) or two other Republican appointees (26.2% liberal votes). A
similar pattern holds true for Democratic appointees.
Table 2: Voting of Federal Court of Appeals Judges in
Sex Discrimination Cases, 1995–2002, by Party of
Appointing President and Panel Colleagues
Republican Appointees Democratic Appointees
Panel
Colleagues
Number of
Observations
Percent
Liberal
Votes
Panel
Colleagues
Number of
Observations
Percent
Liberal
Votes
DD 199 44.2% RR 354 41.5%
RD 708 37.7% DR 398 51.5%
RR 558 26.2% DD 144 78.5%
All cases 1465 34.2% All cases 896 51.9%
Of critical importance, the different outcomes across panel composition
seen in Table 1 do not reflect only simple majoritarian voting.
If judges naïvely voted their policy preferences and case outcomes
were determined by majority vote, then judges would exhibit a stable
voting pattern regardless of the identity of their panel colleagues. As
Table 2 clearly shows, this is not the case. Alternatively, one might
expect that a judge in the ideological minority might be influenced by
her colleagues, but that the two judges in the ideological majority
would not. After all, the majority has the votes to achieve its policy
goals directly. Once again, however, this is not the case; judges in the
ideological majority are also observed to vote differently when a judge
affiliated with the opposing party is on the panel. Thus, the phenomenon
of “panel effects” encompasses two distinct effects: first, that
judges in the majority vote differently (in a less stereotypically ideo-
2009] Empirical Exploration of Panel Effects 1331
logical fashion) than judges on a homogeneous panel; and second,
that judges in the minority vote differently (still less stereotypically
ideologically) than judges in the majority.
B. Theoretical Accounts
What accounts for these observed panel effects? Scholars have
proposed a variety of explanations, encompassing cultural, psychological,
institutional, and strategic factors. In order to frame an empirical
test, I group these explanations into three basic types. As a caveat,
I do not mean to argue that this typology is canonical in any
sense, and each type of explanation that I identify encompasses a
number of diverse theories. Rather than definitively categorizing theories,
this typology merely serves to sharpen the empirical inquiry here.
One type of explanation focuses on the relatively low dissent rates
in court of appeals decisions. A simple ideological model of voting
would predict frequent dissents whenever a panel of judges is divided
ideologically. In fact, the proportion of federal appellate decisions
containing dissents is quite low—around 10% overall.
32
Some scholars
explain the high levels of unanimity by positing the importance of a
“norm of consensus.”
33
Frequent dissents are thought to undermine
institutional legitimacy and the clarity of legal rules,
34
while unanimous
decisions “promote the appearance of legal objectivity, certainty,
and neutrality”
35
and encourage compliance with the law.
36
Other scholars emphasize the costliness of dissent to the individual
judge.
37
Writing a dissenting opinion requires time and effort, and it
32
See HETTINGER ET AL., COLLEGIAL COURT, supra note 21, at 47 (noting that 9.5%
of cases in the U.S. courts of appeals from 1960 to 1996 had dissents).
33
For an overview, see, for example, CROSS, supra note 12, at 160, Burton M. Atkins,
Judicial Behavior and Tendencies Towards Conformity in a Three Member Small Group: A
Case Study of Dissent Behavior on the U.S. Court of Appeals, 54 SOC. SCI. Q. 41, 42-43
(1973), and Farhang & Wawro, supra note 12, at 307 and sources cited therein.
34
See HETTINGER ET AL., COLLEGIAL COURT, supra note 21, at 19 (claiming that
unanimous decisions “may promote institutional legitimacy and effective implementation
of individual decisions”); Edwards, supra note 3, at 1651 (“What the parties and
the public need is [the best] answer, not a public colloquy among judges.”).
35
Farhang & Wawro, supra note 12, at 307.
36
See HETTINGER ET AL., COLLEGIAL COURT, supra note 21, at 19-20 (describing
how “consensual decision making promotes the efficient administration of justice”).
37
See, e.g., CROSS, supra note 12, at 160-61 (explaining why the decision not to dissent
may be a practical response to the costs of dissenting); SUNSTEIN ET AL., ARE
JUDGES POLITICAL?, supra note 10, at 64-66 (describing dissents on three-judge panels
as “both futile and highly burdensome to produce”); Revesz, supra note 10, at 1733
1332 University of Pennsylvania Law Review [Vol. 157: 1319
may negatively impact a judge’s reputation and collegial relations,
38
while offering very little payoff. A dissent has no substantive effect on
the outcome of a case, at least in the short term, and writing one does
not relieve a judge of her responsibilities for drafting opinions in
other cases. These types of theories offer strong reasons that a judge
in the ideological minority might often suppress her disagreement
and go along with the decision of the majority.
Although these theories of “suppressed dissent” offer a plausible
account of why dissents are relatively infrequent on the courts of appeals,
they cannot explain panel effects more generally. As noted
above, panel composition influences not only the behavior of the minority
judge, but the behavior of the judges who comprise the panel
majority as well.
39
As Revesz argued, if judges go along with their colleagues
simply to avoid writing a dissent, one would predict that on
mixed panels, “the single judge of one party [would be] the only one
to moderate his or her views.”
40
The costs of writing a dissent might lead a minority judge to avoid
openly expressing her disagreement, but should have no impact on
the votes of the panel majority. Similarly, a norm of consensus has
more explanatory power for minority than for majority judges. Such a
norm might sometimes induce the majority to accommodate the views
of the minority, but it seems more likely to lead them to ignore the
preferences of the minority, knowing that the strong norm of unanimity
will pressure the minority member to go along.
41
Thus, while theories
of dissent suppression are certainly relevant, they are insufficient
to explain the observed influence of panel composition on the behavior
of both minority and majority judges on mixed panels.
(suggesting that a judge may “moderate[] his or her views in order to avoid having to
write a dissent”).
38
See SUNSTEIN ET AL., ARE JUDGES POLITICAL?, supra note 10, at 66 (“[D]issenting
opinions might also cause a degree of tension among judges . . . .”). Dissents force the
majority judges to confront public disagreement with their conclusions and may oblige
them to respond to arguments raised by the dissent or to more carefully defend the
conclusions that they reach. See Ginsburg & Falk, supra note 4, at 1017 (“Even one dissident
judge can impose upon me the cost, in time and aggravation, of having to respond
to a dissenting opinion . . . .”).
39
See supra Section I.A.
40
Revesz, supra note 10, at 1734.
41
As Sunstein et al. point out, “a Democratic majority, or a Republican majority,
has enough votes to do what it wishes.” SUNSTEIN ET AL., ARE JUDGES POLITICAL?, supra
note 10, at 12.
2009] Empirical Exploration of Panel Effects 1333
The second type of explanation—what I call “internal deliberative”
explanations—focuses on dynamics internal to the judicial panel.
One such explanation is that panel effects are the product of collegial
interactions among appellate judges. This explanation is consistent
with the way that many judges describe the decision-making process
and has been advanced most forcefully by Judge Harry Edwards. He
writes that “if panel composition turns out to have a ‘moderating’ effect
on judges’ voting behavior, this is a sign that panel members
are behaving collegially.”
42
***
Sunstein and his coauthors propose another set of explanations
that focuses on internal panel dynamics—explanations rooted in the
findings of experimental psychology. They cite studies documenting a
“conformity effect,” where individuals in experimental settings are observed
to yield their views in the face of unanimous group opinion to
the contrary,
46
and argue that “judges are vulnerable to similar influ-
ences.”
47
Analogizing the minority judge to the experimental subject
confronted with a unanimous group opinion, they argue that the tendency
to conform to dominant opinion explains why dissents are
far less common on the courts of appeals than a naïve ideological
model would predict.
***
42
Harry T. Edwards, Essay, Collegiality and Decision Making on the D.C. Circuit, 84 VA.
L. REV. 1335, 1358 (1998).
46
SUNSTEIN ET AL., ARE JUDGES POLITICAL?, supra note 10, at 67. As some of those
authors explain in a different work, “[t]he yielding . . . occurs partly because of the information
suggested by the unanimity of others; how could shared views be wrong?
And it occurs partly because of reputational pressures; people do not want to stand out
on a limb for fear that others will disapprove of them.” Sunstein et al., Ideological Voting,
supra note 10, at 339.
47
SUNSTEIN ET AL., ARE JUDGES POLITICAL?, supra note 10, at 69.
1334 University of Pennsylvania Law Review [Vol. 157: 1319
In contrast to dissent-suppression theories and internaldeliberative
accounts, strategic explanations focus on interactions between
the appellate judges on a panel and the other actors in the judicial
system in order to explain panel effects. These accounts posit
that appellate judges do not pursue their policy goals naïvely, but
rather act strategically, with an eye to the likely response of the Supreme
Court or the court of appeals en banc. For example, Virginia
Hettinger, Stefanie Lindquist, and Wendy Martinek propose a strategic
explanation of when appellate judges dissent. They hypothesize
that circuit judges “may choose to dissent to signal the circuit en banc
that the majority panel opinion is contrary to circuit law or contrary to
the preferences of the circuit majority,” or “to signal the Supreme
Court and thereby invite review by that body.”
53
As they recognize, dissenting opinions might also be suppressed if
circuit judges who disagree with the majority opinions nevertheless
believe that en banc or Supreme Court review will produce an outcome
even worse—from their perspective—than the panel majority
53
HETTINGER ET AL., COLLEGIAL COURT, supra note 21, at 41; see also CROSS, supra
note 12, at 156.
opinion. Thus, any prediction about whether or not a circuit judge
will dissent “will depend on the configuration of preferences across the
relevant actors: the judge, the three-judge panel, and the circuit [or the
Supreme Court] as a whole.” Their theory, however, focuses narrowly
on the decision to dissent, rather than on panel effects generally.
***
2009] Empirical Exploration of Panel Effects 1335
Cross and Tiller offer a closely related theory of how strategic behavior
produces observed panel effects. Similar to Hettinger and her
coauthors, they assume that appellate judges use dissents as a signal to
the Supreme Court or the circuit en banc. However, they focus not
on accounting for dissenting behavior, but on explaining why lower
court judges obey precedent.
56
Following doctrine poses no difficulties
where it leads to a result consistent with a circuit judge’s preferences.
However, when existing doctrine does not coincide with her
policy goals, she may be tempted to disregard it. In such a situation,
Cross and Tiller theorize that a panel member who differs ideologically
from the majority will act as a “whistleblower.” By dissenting, the
minority judge can “expose the majority’s manipulation or disregard
of the applicable legal doctrine,”
57
alerting a higher court to the disobedient
decision making and leading to reversal of the original majority
opinion. Alternatively, the threat to “expose disobedient decisionmaking
by the majority” may cause the majority to acknowledge its
“disregard” of doctrine and decide to “keep its decision within
the confines of doctrine.”
58
Cross and Tiller therefore predict that
“courts are more likely to comply with doctrine . . . when the judicial panel
is politically or ideologically divided.”
56
Cross & Tiller, supra note 10, at 2156.
57
Id.
58
Id. at 2159. Judge Wald has expressed skepticism about this account based on
her experience as a judge on the Court of Appeals for the D.C. Circuit: “[T]hreats of
dissent are not particularly effective in changing a panel’s course.” Patricia M. Wald, A
Response to Tiller and Cross, 99 COLUM. L. REV. 235, 253 (1999). Judge Harry Edwards
has been even more blunt, claiming that “the hypothesis is absurd.” Edwards, supra
note 42, at 1337.
Cross and Tiller also suggest an alternative, psychologically based account of whistleblower
effects:
Judges employ “cognitive shortcuts to process imperfect information” under
the legal model, and these shortcuts produce apparently political results. . . .
[T]he minority judge can serve as a whistleblower by revealing these biasing
cognitive shortcuts. Once the majority can no longer readily rationalize its
decision under the legal model, it will frequently concede to the commands of
that model.
1336 University of Pennsylvania Law Review [Vol. 157: 1319
This whistleblowing theory is consistent with models that positive
political theorists commonly use to describe the judicial hierarchy.
60
Briefly, these models analogize the relationship between the Supreme
Court and the lower federal courts to a principal-agent relationship.
The Supreme Court creates doctrine that their “agents,” the lower
federal courts, are supposed to apply faithfully. However, lower court
judges have their own preferences and may be tempted to deviate
from established doctrine. Principal-agent models are thus centrally
concerned with questions of supervision and control—that is, “[h]ow
and to what extent can the Supreme Court control the behavior of
lower federal courts to ensure that its policy dictates are imple-
mented?”
61
One common answer is that lower federal court judges
follow Supreme Court doctrine because they “fear exposure of any
noncompliance and consequent reversal.”
62
The Supreme Court,
however, only reviews a tiny fraction of court of appeals decisions—
currently less than 1% per year.
63
Cross and Tiller’s whistleblowing
Cross & Tiller, supra note 10, at 2174. Sunstein et al. use the whistleblower terminology
in this second sense—as a psychological rather than strategic explanation. SUNSTEIN
ET AL., ARE JUDGES POLITICAL?, supra note 10, at 78-79.
60
For a more detailed discussion of principal-agent models of the federal judicial
hierarchy, see Kim, supra note 11, at 391-404.
61
Id. at 393.
62
Cross & Tiller, supra note 10, at 2158; see also George & Yoon, supra note 24, at
822-25 (noting that the Supreme Court’s mechanism of control over lower courts is its
power of reversal); McNollgast, supra note 22, at 1635-36 (modeling lower court judges
“as strategic actors facing a trade-off between pursuing a personal policy agenda and
seeing their decisions reversed by a higher court”); Songer, Segal & Cameron, supra
note 22, at 693 (“If an appeals court anticipates that it will be sanctioned in the form of
a reversal, the anticipated response will keep the court in check.”).
63
See Kim, supra note 11, at 397-98. Scholars have suggested various mechanisms
by which even a low rate of reversal might induce compliance. For example, Songer,
Segal, and Cameron hypothesize that litigant policing plays a crucial role, suggesting
that losing parties are more likely to petition for Supreme Court review when the lower
court opinion is “noncompliant,” thereby sounding a “fire alarm” that alerts the Court
to cases of “flagrant doctrinal shirking.” Songer, Segal & Cameron, supra note 22, at
693. McNollgast argue that the Supreme Court exercises effective control by establishing
a “doctrinal interval” of acceptable outcomes in order to induce lower courts to
follow its precedents. McNollgast, supra note 22, at 1645-46. These explanations have
been criticized on theoretical grounds, and the handful of relevant empirical studies
generally do not support the theory that fear of reversal motivates lower court compliance
with doctrine. See Kim, supra note 11, at 399-404 and sources cited therein; see
also CROSS, supra note 12, at 99-101.
2009] Empirical Exploration of Panel Effects 1337
theory offers one possible mechanism by which the Supreme Court
might efficiently monitor and control the decisions of the courts of
appeals—relying on dissenting opinions to signal cases of noncompliance
that warrant review.
Theorists have similarly analogized the relationship between a circuit
court and its three-judge panels to an agency relationship.
64
On
this view, individual judges are not free to decide as they like, but must
act as “representatives” of the circuit.
65
A three-judge panel is “deputed
to hear and to determine cases in conformity with the law as the
full court views it.”
66
To ensure that this representative function is carried
out faithfully, the majority of the full circuit is permitted to overrule
a panel decision by rehearing a case en banc.
67
Like the Supreme
Court, however, the circuit as a whole will find it costly to monitor the
decisions of each panel. To solve this monitoring problem, the circuit
may rely on signals such as the presence of a dissenting opinion to determine
which cases to rehear en banc, and circuit court judges, aware
of this possibility, may vote strategically in order to invite or avoid en
banc review of a panel’s decision.
68
Both Hettinger et al.’s strategic-dissent theory and Cross and
Tiller’s whistleblower theory draw some support from the fact that the
presence of a dissenting opinion is associated with both a greater likelihood
that a case will be reheard en banc
69
and that the Supreme
Court will grant certiorari.
70
However, this observed correlation does
not necessarily prove that dissenting opinions cause the circuit en banc
or the Supreme Court to review a case. It may be that both the existence
of a dissent and the decision to rehear or accept certiorari are
the result of some underlying characteristic of the case—for example,
64
E.g., George, supra note 26, at 245; Ginsburg & Falk, supra note 4, at 1011-13;
Michael E. Solimine, supra note 4, at 49 (1988).
65
Solimine, supra note 4, at 49.
66
Ginsburg & Falk, supra note 4, at 1011.
67
Solimine, supra note 4, at 49.
68
Litigants may also play a role in monitoring panel decisions for the circuit, because
they are more likely to petition for rehearing en banc if they believe that a panel
decision is contrary to the preferences of the circuit majority. See Michael W. Giles,
Thomas G. Walker & Christopher Zorn, Setting a Judicial Agenda: The Decision to Grant
En Banc Review in the U.S. Courts of Appeals, 68 J. POL. 852, 865 (2006) (presenting empirical
evidence that litigants’ decisions to seek en banc review are influenced by the
ideological preferences of the panel relative to those of the circuit majority).
69
George, supra note 26, at 267; Douglas H. Ginsburg & Brian M. Boynton, The
Court En Banc: 1991–2002, 70 GEO. WASH. L. REV. 259, 264 (2002); Ginsburg & Falk,
supra note 4, at 1046.
70
Caldeira, Wright & Zorn, supra note 22, at 563 tbl.1.
1338 University of Pennsylvania Law Review [Vol. 157: 1319
that it involves a particularly difficult or close legal issue. Moreover,
even if the relationship between dissents and further review is a causal
one, a considerable gap remains between the large number of court
of appeals cases containing dissents and the very limited number accepted
for Supreme Court or en banc review.
71
Thus, while the presence
of a dissent may encourage the Supreme Court or circuit en
banc to hear a case, it remains uncertain whether the possibility of a
dissent and subsequent review actually influences the panel behavior
of court of appeals judges.
71
As discussed above, dissents occur in about 10% of federal appellate cases. See
supra note 32 and accompanying text. In contrast, the probability that a court of appeals
decision will be reviewed by the Supreme Court and the probability of review by
the circuit en banc are quite small, with both events occurring in less than 1% of cases.
See Kim, supra note 11, at 391 n.30 (“[T]he chance that a given court of appeals decision
will be reviewed by the Supreme Court is approximately 0.14%.”); see also Michael
W. Giles, Virginia A. Hettinger, Christopher Zorn & Todd C. Peppers, The Etiology of the
Occurrence of En Banc Review in the U.S. Courts of Appeals, 51 AM. J. POL. SCI. 449, 450
(2007) (explaining that while the incidence of en banc hearings varies significantly
across circuits and across time, the incidence is “uniformly low”); Ginsburg & Boynton,
supra note 69, at 266 tbl.6 (reporting that the percentage of cases heard en banc from
1997 to 1999 varied between 0.10% and 0.58% depending upon the circuit); Ginsburg
& Falk, supra note 4, at 1045 tbl.5 (reporting that 1.03% of argued cases and 0.2% of
nonargued cases were reheard en banc by the D.C. Circuit from 1981 to 1990);
Solimine, supra note 4, at 46 tbl.2 (reporting that less than 1% of court of appeals cases
were heard en banc in the 1980s).
One might argue that the low percentage of cases actually reviewed by the Supreme
Court or the circuit en banc does not necessarily indicate a lack of control by
the reviewing courts, but rather the rate of review is low because control is effective
and extensive oversight is not necessary. Thus, the actual rate of review cannot establish
the true level of effective control.
If, however, a reviewing court is relying on the threat of review to exercise control
over panel decisions, that threat must be at least a credible one. Cross describes the
problem in the context of the Supreme Court:
[T]here surely must be some credible threat of admonishment to maintain
control on the playground. It is doubtful that nine teachers (the Supreme
Court justices), who are capable of admonishing at most around one hundred
students a year, could effectively keep order on a playground populated by
more than fifty thousand students.
CROSS, supra note 12, at 100.
Given resource constraints, neither the Supreme Court nor the circuits have the
capacity to increase significantly the proportion of panel decisions reviewed. Thus,
while the threat of review might sometimes make actual review unnecessary, the limited
capacity of the reviewing courts suggests that fear of reversal may not play a dominant
role in appellate panel decision making.