Electronic copy available at: http://ssrn.com/abstract=2087255
LEGAL STUDIES RESEARCH PAPER SERIES
PAPER NO. 12-05-31
May 2012
DOES PUBLIC OPINION INFLUENCE THE SUPREME COURT?
POSSIBLY YES (BUT WE’RE NOT SURE WHY)
by
Lee Epstein
Provost Professor of Law and Political Science
Rader Family Trustee Chair in Law
University of Southern California
and
Andrew D. Martin
Professor of Law and Political Science
Electronic copy available at: http://ssrn.com/abstract=2087255
263
DOES PUBLIC OPINION INFLUENCE THE
SUPREME COURT? POSSIBLY YES
(BUT WE’RE NOT SURE WHY)
Lee Epstein
*
Andrew D. Martin
**
INTRODUCTION
Using qualitative data and historical methods, Barry Friedman asserts
with confidence that “we the people” influence the decisions of
the U.S. Supreme Court.
1
Using quantitative data and statistical methods,
political scientists are not so sure. Despite their best efforts to
validate basic claims about the effect of public opinion on the Court,
the evidence remains mixed at best.
2
We enter this dialogue but in a voice distinct from existing political
science work. Rather than explore the relationship between the
public and the Court on a term-by-term basis, we analyze it at the level
of the case. This allows us to exploit more nuanced public opinion
data, as well as to attend to the many other case-level factors that may
influence the Court’s decisions.
Based on our analysis, we are prepared to say that Professor
Friedman is on to something. When the “mood of the public” is liberal
(conservative), the Court is significantly more likely to issue liberal
(conservative) decisions. But why is anyone’s guess. Professor
* Lee Epstein (http://epstein.law.northwestern.edu/) is the Henry Wade Rogers Professor
at Northwestern.
** Andrew D. Martin (http://adm.wustl.edu/) is Chair of the Political Science Department
and Professor of Law and Political Science at Washington University in St. Louis. We are
grateful to Barry Friedman, Linda Greenhouse, and other participants at the 2010 Journal
of Constitutional Law’s Symposium for their helpful comments. We also thank Jeffrey A.
Segal and Richard A. Posner for their insights, the National Science Foundation for research
support, and Jeremy Caddel and Jee Seon Jeon for research assistance. The
project’s website houses a full replication archive: http://epstein.law.northwestern.edu/
research/PublicOpinion.html.
1 The full title of Professor Friedman’s book tells the story. BARRY FRIEDMAN, THE WILL OF
THE PEOPLE: HOW PUBLIC OPINION HAS INFLUENCED THE SUPREME COURT AND SHAPED
THE MEANING OF THE CONSTITUTION 14 (2009) (stating that the decisions of the Supreme
Court on contentious issues align with popular approval and public understanding of the
Constitution).
2 See infra Part II.
Electronic copy available at: http://ssrn.com/abstract=2087255
264 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
Friedman posits that the Justices will bend to the will of the people
because the Court requires public support to remain an efficacious
branch of government. Our analysis could be read to support this
view, but it is equally consistent with another mechanism: that “the
people” include the Justices. On this account, the Justices do not respond
to public opinion directly, but rather respond to the same
events or forces that affect the opinion of other members of the pub-
lic.
3
Or, as Cardozo once stated, “[t]he great tides and currents
which engulf the rest of men do not turn aside in their course and
pass the judges by.”
4
Our study proceeds as follows. In Part II, we briefly review the extant
literature, emphasizing the similar methodology it invokes but
the varying conclusions it reaches. Parts III and IV describe our methods
and findings. We end, in Part V, with the implications of our
statistical work for Professor Friedman’s claims, as well as for future
research assessing the Court’s response to public opinion.
II. THE EXISTING LITERATURE
Professor Friedman’s book is only the last in a long and distinguished
line of research evaluating the link between the public and
the courts. Much, perhaps most, of the work in the social sciences
has explored the flipside of Professor Friedman’s interest: the extent
to which Court decisions affect (or “move”) public opinion.
5
But
3 The debate between these two accounts has a long history in the political science literature.
Compare Beverly B. Cook, Public Opinion and Federal Judicial Policy, 21 AM. J. POL. SCI.
567, 574 (1977) (showing that federal judges were more lenient with Vietnam War draft
offenders as public opinion toward the war grew increasingly negative), with Herbert M.
Kritzer, Federal Judges and Their Political Environments: The Influence of Public Opinion, 23
AM. J. POL. SCI. 194, 195–98 (1979) (showing that changes in severity of sentencing did
not move in the same way as changes in public opinion, possibly reflecting “the judges’
own doubts about the war” rather than a direct response to public opinion). See generally
JEFFREY A. SEGAL & HAROLD J. SPAETH, THE SUPREME COURT AND THE ATTITUDINAL
MODEL REVISITED 424–25 (2002) (explaining that while it is unsurprising that Supreme
Court decisions generally correspond with public opinion, it is difficult to determine
whether public opinion directly influences the Court).
4 BENJAMIN N. CARDOZO, THE NATURE OF THE JUDICIAL PROCESS 168 (1921).
5 The classic work is Charles H. Franklin & Liane C. Kosaki, Republican Schoolmaster: The
U.S. Supreme Court, Public Opinion, and Abortion, 83 AM. POL. SCI. REV. 751, 768 (1989)
(finding that the Court’s decisions have an impact on the public, but that the impact manifests
itself as both increased support and increased opposition for the position taken by
the Court’s majority). For recent research, see, e.g., Brandon L. Bartels & Diana C. Mutz,
Explaining Processes of Institutional Opinion Leadership, 71 J. POL. 249, 259 (2009), reporting
experimental evidence that an endorsement from the Court affected respondents’ opinions
about particular public policy issues; Mark D. Ramirez, Procedural Perceptions and
Support for the U.S. Supreme Court, 29 POL. PSYCHOL. 675, 676 (2008), arguing that positive
Dec. 2010] WE’RE NOT SURE WHY 265
there is no shortage of quantitative (and multivariate) work addressing
directly the question of whether the public influences the Court.
Rather than waste space with lengthy descriptions of each paper, we
summarize the key results in Table 1. We limit our list to studies of
the U.S. Supreme Court, the focus of Professor Friedman’s book.
6
TABLE 1
SUMMARY OF NOTABLE MULTIVARIATE STUDIES ON THE EFFECT OF
PUBLIC OPINION ON THE DECISIONS OF THE SUPREME COURT OR
INDIVIDUAL JUSTICES
Author (Year) Description of Methods and Key Findings
Clark (2009)7
Examines the number of federal laws struck down
by the Court each term; finds that as public support for
the Supreme Court declines, the Court strikes down
fewer laws.
media coverage increases support for the Court; James W. Stoutenborough, and Donald
P. Haider-Markel & Mahalley D. Allen, Reassessing the Impact of Supreme Court Decisions on
Public Opinion: Gay Civil Rights Cases, 59 POL. RES. Q. 419, 430 (2006), asserting that the
Court is able to legitimate controversial policies. See also Valerie J. Hoekstra, The Supreme
Court and Local Public Opinion, 94 AM. POL. SCI. REV. 89, 91, 97 (2000) (relaying the ability
of particular Court decisions to affect overall evaluations of the Court in the geographic
areas that gave rise to the case).
6 Studies of other federal judges include Cook, supra note 3, at 574; Kritzer, supra note 3, at
195–98. There are also studies of state court judges. These are more uniform in their results,
tending to confirm the obvious: that (elected) state court judges must consider
their constituents’ preferences to keep their jobs. See, e.g., Steven P. Croley, The Majoritarian
Difficulty: Elective Judiciaries and the Rule of Law, 62 U. CHI. L. REV. 689, 696–97, 788
(1995) (criticizing elective judiciaries because they respond to majoritarian political pressures);
James H. Kuklinski & John E. Stanga, Political Participation and Government Responsiveness:
The Behavior of California Superior Courts, 73 AM. POL. SCI. REV. 1090, 1093 (1979)
(explaining findings that California superior courts less in line with public opinion
showed adjustments in their decisions); Alexander Tabarrok & Eric Helland, Court Politics:
The Political Economy of Tort Awards, 42 J.L. & ECON. 157, 187 (1999) (indicating that
elected judges rule according to popular views because they wish to be re-elected). Finally,
a growing number of scholars are exploring the relationship between public sentiment
and judicial decisions in courts outside the United States. Along these lines, we commend
JEFFREY K. STATON, JUDICIAL POWER AND STRATEGIC COMMUNICATION IN MEXICO 3–
4 (2010). Staton documents the Mexican Supreme Court’s “coordinated and aggressive”
public relations campaign to create public support. He also shows that courts throughout
the world publicize their decisions, issue press releases, and maintain websites describing
their procedures and publishing their cases and biographies of their members.
7 Tom S. Clark, The Separation of Powers, Court Curbing, and Judicial Legitimacy, 53 AM. J. POL.
SCI. 971, 985 (2009).
266 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
Giles et al. (2008)8
McGuire & Stimson
(2004)9
Explores the percentage of liberal votes cast by each
Justice each term; finds that public opinion exerts only
a small effect and that it operates on “as few as 20%” of
the Justices serving between 1956 and 1999.
Analyzes the percentage of liberal decisions each
term in cases in which the Court reverses the court below;
shows that the public’s mood affects the percentage
even after controlling for ideology.
Flemming &
Wood (1997)10
Analyzes the percentage of liberal votes cast by each
Justice each term; shows that the Court’s composition
accounts for most of the variation, but that the public’s
mood also exerts a small (though significant and quick)
effect on most Justices in the majority of legal areas under
analysis.
Mishler & Sheehan
(1996)11
Considers the percentage of liberal votes cast by
each Justice each term; finds that some moderate Justices
respond to the public mood and concludes that it
“shapes the subjective norms of individual justices thereby
potentially influencing their decisions.”12
Stimson et al.
(1995)13
Analyzes the ideological direction of decisions each
term; unearths some responsiveness to the mood of the
public, but the effect nearly disappears when controlling
for the Court’s composition.
Link (1995)14
Examines the percentage of liberal decisions each
8 Micheal W. Giles, Bethany Blackstone & Richard L. Vining, Jr., The Supreme Court in American
Democracy: Unraveling the Linkages Between Public Opinion and Judicial Decision Making,
70 J. POL. 293, 296, 303 (2008).
9 Kevin T. McGuire & James A. Stimson, The Least Dangerous Branch Revisited: New Evidence
on Supreme Court Responsiveness to Public Preferences, 66 J. POL. 1018, 1033 (2004).
10 Roy B. Flemming & B. Dan Wood, The Public and the Supreme Court: Individual Justice Responsiveness
to American Policy Moods, 41 AM. J. POL. SCI. 468, 474, 482 (1997).
11 William Mishler & Reginald S. Sheehan, Public Opinion, the Attitudinal Model, and Supreme
Court Decision Making: A Micro-Analytic Perspective, 58 J. POL. 169, 180 (1996).
12 Id. at 174.
13 James A. Stimson, Michael B. MacKuen & Robert S. Erikson, Dynamic Representation, 89
AM. POL. SCI. REV. 543, 552, 556 (1995).
Dec. 2010] WE’RE NOT SURE WHY 267
term in criminal procedure and race cases; finds that
the public’s mood, to greater and lesser extents (and
with some lags), has a direct effect on Court decisions.
Norpoth & Segal
(1994)15
Reexamines Mishler & Sheehan’s 1993 study; finds
that any change in the direction of the Court’s decisions
was due to an “indirect model” of public influence
in which “popularly elected presidents, through
new appointments, affect the ideological composition
of the Court.”16
Mishler &
Sheehan (1993)17
Explores the percentage of liberal decisions each
term; finds that (lagged) public mood influences
Court’s decisions both indirectly (through composition
changes) and directly; claims that the Justices “are
broadly aware of fundamental trends in the ideological
tenor of public opinion, and . . . at least some justices,
consciously or not, may adjust their decisions at the
margins to accommodate such fundamental trends.”18
Two features of the studies in Table 1 stand out: their common
methodology and their mixed findings. Both deserve some elabora-
tion.
A. Common Methods
As even a cursory glance at Table 1 reveals, the studies tend to follow
the same methodological approach to answer the question of
whether public opinion influences the Court. They begin by calculating
the percentage of liberal decisions (or votes) that the Court (an
individual Justice) reaches each term—or, in the jargon of social
science, by aggregating decisions (votes) on a term-by-term basis.
Figure 1 provides an example.
14 Michael W. Link, Tracking Public Mood in the Supreme Court: Cross-Time Analyses of Criminal
Procedure and Civil Rights Cases, 48 POL. RES. Q. 61, 67, 75 (1995).
15 Helmut Norpoth & Jeffrey A. Segal, Comment, Popular Influence on Supreme Court Decisions,
88 AM. POL. SCI. REV. 711, 711 (1994).
16 Id.
17 William Mishler & Reginald S. Sheehan, The Supreme Court as a Countermajoritarian Institution?
The Impact of Public Opinion on Supreme Court Decisions, 87 AM. POL. SCI. REV. 87, 90,
96 (1993).
18 Id. at 89.
268 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
FIGURE 1. PERCENTAGE OF LIBERAL DECISIONS, 1953–2008 TERMS
1960 1970 1988 1990 2000 2010
Term
Because almost all the authors use the U.S. Supreme Court Database
to categorize decisions as liberal or conservative (as do we in
Figure 1), their definitions of ideology do not vary much. Where differences
emerge are in the types of cases they include in their study.
For example, some studies analyze only a few areas of the law;
19
one
focuses on reversals;
20
and another separates the cases into “salient”
and “non-salient” categories.
21
Figure 1 is a generic depiction of the
percentage, using all cases decided between the 1953–2008 terms.
22
(As an aside, readers should take care in interpreting Figure 1 or any
other display based on “percent liberalism” because the types of cases
the Justices agree to hear and decide in a given term will affect the
percentage.)
19 See, e.g., Link, supra note 14, at 67 (examining only criminal procedure and civil rights
cases). Others analyze separately particular areas of the law or control for areas in the
analysis. See, e.g., Flemming & Wood, supra note 10, at 485–92 (evaluating whether specific
issue areas impacted judicial responsiveness to public opinion); McGuire & Stimson,
supra note 9, at 1025 (reflecting case outcome types for distinct issue areas).
20 McGuire & Stimson, supra note 9, at 1019.
21 Giles, Blackstone & Vining, supra note 8, at 296.
22 The percentage was calculated using the U.S. Supreme Court Database, THE SUPREME
COURT DATABASE, http://supremecourtdatabase.org (last visited Oct. 7, 2010) (cases organized
by citation), and orally argued cases (decisionType = 1 or 6 or 7) in which the direction
of the decision is liberal or conservative (i.e., we omit the 60 cases coded as “un-
specifiable”).
Dec. 2010] WE’RE NOT SURE WHY 269
The percentage of liberal decisions (votes) per term is the dependent
variable in the studies (i.e., what the authors are trying to explain).
The key independent variable is, of course, public opinion.
In other words, the authors are interested in determining whether a
more liberal public leads the Court (or the individual Justices) to
produce more liberal decisions (votes) each term, all else being
equal. To measure public opinion, the studies typically rely on Stimson’s
indicator of “public mood.”
23
Computed by analyzing survey
responses to a range of questions with a sophisticated dynamic factor
analysis model, the indicator is a single number summarizing the
mood of the public along a liberal-conservative dimension. Although
Stimson calculates his measure on a quarterly basis, the articles in
Table 1 use an annual indicator of mood because their dependent
variable (percentage liberalism) is also aggregated (though usually to
the term rather than the year). Figure 2 displays this annual version.
FIGURE 2. STIMSON’S ANNUAL MEASURE OF THE “PUBLIC MOOD,” 1953-
2008
1950 1960 1970 1980 1990 2000 2010
Year
23 JAMES A. STIMSON, PUBLIC OPINION IN AMERICA: MOODS, CYCLES, AND SWINGS (2d ed.
1999). Updated at James A. Stimson, Time Series, UNIV. OF N.C. AT CHAPEL HILL,
http://www.unc.edu/~jstimson/time.html.
270 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
B. Mixed Findings
The final step in most of the studies in Table 1 is to analyze
whether the annual public mood measure explains term-by-term variation
in the Court’s (Justices’) liberal decisions (votes), while controlling
for other factors that also might affect the percentage of liberal
decisions (votes) each term. These other factors almost always include
a measure of the Court’s ideology but otherwise tend to vary
from study to study.
24
Also varying from study to study are the results, though perhaps it
is worth starting with a common finding: virtually all the studies
demonstrate an indirect effect of public opinion via the appointments
process. That is, they tend to show that changes in the Court’s composition
can lead to a higher or lower percentage of liberal decisions
each term. Political scientists deem this “indirect” because the public
does not directly affect the percentage of liberal decisions; its role
comes in electing the President and the Senate, who appoint and
confirm Justices reflecting the public’s preferences.
25
Accordingly,
“the ideological orientation of the Court generally corresponds to the
attitudes of the electorate”
26
and the ruling regime.
The more controversial matter is whether the public directly influences
Court decisions (as Friedman claims). The answers from the
existing literature, as Table 1 suggests, run the gamut. Perhaps the
strongest supporters of a direct, causal link between public mood and
Court decisions, McGuire and Stimson, are as unequivocal as Friedman:
“[P]ublic opinion is a powerful influence on the decisions of
the Supreme Court. . . . [The evidence shows that] a system of popular
representation is alive and well in the Supreme Court.”
27
In direct
juxtaposition comes Norpoth and Segal’s famous reanalysis of Mishler
& Sheehan’s 1993 study, in which they declare that “contrary to
24 E.g., Flemming & Wood, supra note 10, at 468 (controlling for different areas of the law);
Giles, Blackstone & Vining, supra note 8, at 296 (distinguishing salient and non-salient
cases); Link, supra note 14, at 69–70 (controlling for the preferences of Congress and the
President).
25 Justice Antonin Scalia put it this way: “Ultimately, the [J]ustices of the Court are taken
from the society, . . . and however impartial they may try to be, they are going to bring
with them those societ[al] attitudes.” DAVID M. O’BRIEN, STORM CENTER: THE SUPREME
COURT IN AMERICAN POLITICS 343 (5th ed. 2000) (alteration in original).
26 McGuire & Stimson, supra note 9, at 1020; see also Robert A. Dahl, Decision-Making in a
Democracy: The Supreme Court as a National Policy-Maker, 6 J. PUB. L. 279, 293 (1957) (arguing
that the Supreme Court supports the major policies of the dominant political power
because doing otherwise risks undermining its own legitimacy if its rulings are not en-
forced).
27 McGuire & Stimson, supra note 9, at 1033.
Dec. 2010] WE’RE NOT SURE WHY 271
Mishler and Sheehan, [we find] no evidence for a direct path of influence
from public opinion to Court decisions.”
28
Charting a more
moderate course are those scholars who identify some “direct” but
relatively small effect. Stimson et al., fall into this category:
Norpoth and Segal (1994) argue that the role of public opinion on
Court decisions is wholly indirect through the election-nominationconfirmation
process. Mishler and Sheehan (1993, 1994) claim a direct
public opinion influence. Our results, with similar data but quite different
model specification, leave us in the middle of this debate. We believe
there is a trace of influence for both processes, but our results are too
weak to leave us confident about the matter.
29
III. OUR STUDY
Although the studies listed in Table 1 are thoroughly professional
efforts, published in top disciplinary journals, features of their research
design raise some red flags. Of particular concern is that they
cannot incorporate controls at the case level because they aggregate
by term. To provide but one example: we know that the Justices
tend to reverse the decisions of the court below (usually a U.S. court
of appeals).
30
To account for this tendency, researchers modeling the
ideological direction of the Court’s decisions (liberal or conservative)
almost always incorporate a variable encoding the direction of the
lower court’s decision (liberal or conservative).
31
The idea is that
even a very liberal (conservative) Supreme Court would be inclined
to reverse a liberal (conservative) lower court decision given the majority’s
general propensity to reverse.
It is easy to add this variable if the study models decisions on a
case-by-case basis; it is impossible to do so if the study models decisions
on a term-by-term basis—as the existing public opinion research
does—without running the risk of ecological fallacy. The same holds
for many other covariates of Court decisions. Although some are annual
(such as the ideology of the median Justice), others are not, including
the issue area and the parties. Aggregating by term is not
28 Norpoth & Segal, supra note 15, at 711.
29 Stimson, MacKuen & Erikson, supra note 13, at 556.
30 Between the 1953 and 2008 terms, the Court ruled for the petitioner in 64% of the 6286
orally argued cases in which the Supreme Court Database specifies the direction of the
decision and a winning party. THE SUPREME COURT DATABASE, supra note 22.
31 See, e.g., SEGAL & SPAETH, supra note 3; Lee Epstein et al., Circuit Effects: How the Norm of
Federal Judicial Experience Biases the Supreme Court, 157 U. PA. L. REV. 833, 872 (2009) (“controlling
for the ideological direction of the lower court’s decision” in an analysis of the effects
of Justices’ circuit court experience).
272 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
even an especially good strategy for considering the effect of the public’s
mood on the Court because it fails to exploit Stimson’s quarterly
data.
For these reasons, our analysis of Friedman’s thesis proceeds at
the case level. The dependent variable is the ideological direction of
each decision in all orally argued cases between 1958 and 2008
32
—
and not the percentage liberal each term. That is, we intend to study
the decisions in 5,675 cases, rather than a single percentage in 50
terms.
By “ideological direction,” we mean the direction of the decision
(liberal or conservative), as coded by the U.S. Supreme Court Database,
the same source most of the other public opinion studies use.
33
The database’s classifications generally comport with conventional
understandings. “Liberal” decisions are those in favor of defendants
in criminal cases; of women and minorities in civil rights cases; of individuals
against the government in First Amendment, privacy, and
due process cases; of unions over individuals and individuals over
businesses in labor cases; and of the government over businesses in
cases involving economic regulation. “Conservative” decisions are the
reverse. Table 2 provides basic summary information about this variable
(and all others in our analysis).
TABLE 2
Summary Statistics (N=5675)
Variable Description Mean
(Std.
Dev.)
Dependent Variable
Direction of Decision
Whether the Court reached a liberal
(=1) or conservative (=0)decision
0.501
(0.500)
Key Independent
Variable
Public Mood
Stimson’s quarterly measure of
public mood (the higher the score,
the more liberal the public)
61.06234
(4.628)
32 Quarterly public mood data begin with the fourth quarter of 1958 (covering all but two
cases in the 1958 term) through the fourth quarter of 2008 (covering only four cases of
the 2008 term).
33 THE SUPREME COURT DATABASE, supra note 22.
34 The minimum is 51.363; the maximum is 74.18.
Dec. 2010] WE’RE NOT SURE WHY 273
Control Variables
Direction of the Lower
Court Decision
Whether the court whose decision
the Supreme Court reviewed
reached a liberal (=1) or conservative
(=0) decision
0.463
(0.499)
Ideology of the Supreme
Court
Martin & Quinn’s estimate of the
ideology of the median justice (the
higher the score, the more conservative
the median)
0.36435
(0.495)
U.S. Liberal Party
Whether the United States (or an
agency), as petitioner or respondent,
argued for a liberal decision
(=1) or not (=0)
0.065
(0.247)
U.S. Conservative Party
Whether the United States (or an
agency), as petitioner or respondent,
argued for a conservative decision
(=1) or not (=0)
0.134
(0.340)
Civil Liberties
Civil Liberties case (=1) or not
(=0)36
0.549
(0.498)
Economics Economics case (=1) or not (=0)37
0.275
(0.446)
Judicial Power
Judicial power case (=1) or not (=0)
0.129
(0.335)
Federalism Federalism case (=1) or not (=0)
0.048
(0.214)
Ideology of the Poole’s Common Space score (the -0.21138
35 The minimum is -.85; the maximum is 1.027.
36 The Civil Liberties category includes Criminal Procedure, Civil Rights, First Amendment,
Due Process, Privacy, and Attorneys (i.e., “issueArea” = 1, 2, 3, 4, 5, 6) in the U.S. Supreme
Court Database. THE SUPREME COURT DATABASE, supra note 22.
37 The Economics category includes Unions, Economic Activity, and Taxation (i.e., “issueArea"
= 7, 8, 12) in the U.S. Supreme Court Database. THE SUPREME COURT DATABASE, supra
note 22.
274 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
President higher the score, the more conservative
the President)
(0.519)
Ideology of the House
Poole’s Common Space score (the
higher the score, the more conservative
the House)
0.05739
(0.114)
Ideology of the Senate
Poole’s Common Space score (the
higher the score, the more conservative
the Senate)
0.06940
(0.106)
The independent variable of primary interest is Stimson’s quarterly
public mood measure. Because lower scores are more conservative,
we expect this variable to be positively related to the direction of the
decision. In other words, if Friedman is right and public opinion influences
the Court, then the more liberal the public mood, the higher
the likelihood of a liberal decision.
Finally, we include in the model a host of variables that might also
affect the Court’s decision (see Table 2). Seven are case-level variables:
the direction of the lower court’s decision, the side
represented by the U.S. government (if it was a party to the suit, liberal
or conservative), and four issue dummies (“Federalism” is the
omitted category in the model). Given the Court’s tendency to reverse
(and our coding of the relevant variables), we expect the direction
of the lower court decision to be negatively related to the Court’s
decision. As for the U.S. government, the existing literature suggests
that the side it represents will prevail more often than not.
41
If so,
38 The minimum is -.724; the maximum is .503.
39 The minimum is -.226; the maximum is .194.
40 The minimum is -.199; the maximum is .183.
41 See, e.g., REBECCA MAE SALOKAR, THE SOLICITOR GENERAL: THE POLITICS OF LAW 145
(1992) (“[B]etween 1959 and 1986. . . . [t]he party that the government supported with
an amicus brief won 71.9 percent of the time.”); JEFF YATES, POPULAR JUSTICE:
PRESIDENTIAL PRESTIGE AND EXECUTIVE SUCCESS IN THE SUPREME COURT 96 (John Kenneth
White ed., 2002) (noting the “overwhelmingly successful record” the solicitor general
has before the Court); see also RICHARD L. PACELLE, JR., BETWEEN LAW AND POLITICS:
THE SOLICITOR GENERAL AND THE STRUCTURING OF RACE, GENDER, AND REPRODUCTIVE
RIGHTS LITIGATION 5 (James P. Pfiffner ed., 1st ed. 2003) (“The Court . . . grants a higher
percentage of the solicitor general’s petitions and the government is more successful on
the merits than any other litigant.”); ROBERT SCIGLIANO, THE SUPREME COURT AND THE
PRESIDENCY 177 (Samuel Krislov ed., 1971) (“Throughout its history the United States has
won most of its cases in the Supreme Court.”); Tracey E. George & Lee Epstein, On the
Nature of Supreme Court Decision Making, 86 AM. POL. SCI. REV. 323, 325 (1992) (“[W]e have
several reasons to suspect that Congress and the [P]resident affect Supreme Court decision
making . . . .”); Jeffrey A. Segal & Cheryl Reedy, The Supreme Court and Sex Discrimina-
Dec. 2010] WE’RE NOT SURE WHY 275
“U.S. Liberal Party” should be positively signed; “U.S. Conservative
Party” should be negatively signed.
This leaves the four term-level covariates, all of which attempt to
assess the ideology of the various branches of government. For the
U.S. House of Representatives, the U.S. Senate and the U.S. President,
we use Poole’s Common Space scores. These reflect roll call
votes in each House or, in the case of the President, his position on
bills before Congress.
42
We measure the Court’s ideology using Martin and Quinn’s ideal
point estimate of the median Justice each term.
43
The Martin and
Quinn measure is also based on votes—all those cast by the Justices in
non-unanimous cases. Table 2 supplies summary statistics on all four
measures of ideology; Figure 3 depicts them over time.
FIGURE 3. IDEOLOGY OF THE THREE BRANCHES OF GOVERNMENT
3a. The Senate, House, and President
1960 1970 1980 1990 2000 2010
Year
tion: The Role of the Solicitor General, 41 POL. RES. Q. 553, 555–56 (1988) (“[T]he Supreme
Court is responsive to the executive branch of government through its support for the solicitor
general.”).
42 VOTEVIEW.COM, http://voteview.com/ (last visited Oct. 7, 2010); see also KEITH T. POOLE
& HOWARD ROSENTHAL, CONGRESS: A POLITICAL-ECONOMIC HISTORY OF ROLL CALL
VOTING 7 (1997) (analyzing the dynamics of roll call voting using a liberal/conservative
structure).
43 MARTIN-QUINN SCORES, http://mqscores.wustl.edu/ (last visited Sep. 22, 2010); see Andrew
D. Martin & Kevin M. Quinn, Dynamic Ideal Point Estimation via Markov Chain Monte
Carlo for the U.S. Supreme Court, 1953–1999, 10 POL. ANALYSIS 134, 152 (2002) (investigating
judicial preferences and finding that ideal points of many Justices change over time).
276 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
3b. The Court
1960 1970 1980 1990 2000 2010
Year
Under Poole’s and Martin and Quinn’s coding schemes, the lower
the ideology score, the more liberal the branch. Hence, if these variables
affect the Court’s decision, we expect negatively signed coefficients
(i.e., the lower the score, the higher the likelihood of a liberal
decision).
IV. RESULTS
To assess the effect of the public’s mood on the ideological direction
of the Court’s decision, we use logistic regression (with various robustness
checks).
44
Table 3 displays the results. Note that we cluster
the observations by term, which helps account for any term-to-term
correlations in the data (and results in higher standard errors).
44 For example, we reestimated the model using a random effects logistic regression. The
results are virtually identical to those displayed in Table 3. Most importantly, the coefficient
on Public Mood (Quarterly) remains statistically significant (.023, with a standard error
of .007).
Dec. 2010] WE’RE NOT SURE WHY 277
TABLE 3 RESULTS OF REGRESSING THE DIRECTION OF THE COURT’S
DECISION (1=LIBERAL; 0=CONSERVATIVE) ON THE PUBLIC MOOD AND
OTHER COVARIATES
Variable Coefficient (Standard Er-
ror)
Public Mood (Quarterly) 0.023 (0.008)
Direction of the Lower Court Decision -1.041 (0.070)
Ideology of the Supreme Court -0.547 (0.049)
U.S. Liberal Party 0.316 (0.131)
U.S. Conservative Party -0.757 (0.095)
Civil Liberties -0.287 (0.134)
Economics -0.033 (0.135)
Judicial Power -1.015 (0.148)
Ideology of the President 0.017 (0.056)
Ideology of the House -0.140 (0.331)
Ideology of the Senate 0.283 (0.412)
Intercept -0.367 (0.459)
Note: Robust standard errors reported, clustered by term. Italicized variables are
statistically significant at the =0.05 level. Federalism is the omitted category. N=5675.
Beginning with the primary variable of interest, Public Mood (Quarterly),
we find support for Friedman’s basic claim. The coefficient is
statistically significant and positively signed, indicating that the more
liberal the public mood, the higher the likelihood of a liberal deci-
sion.
But how much higher? Keeping in mind that the Court usually
takes cases to reverse,
45
the predicted probability of a liberal decision
is .59
46
when the Justices review a conservative lower court decision
and the public mood is very conservative.
47
If the public mood is very
liberal, the probability jumps to .71.
48
Alternatively, if the Court reviews
a liberal lower court decision when the public is very conservative,
the probability that it will affirm (i.e., reach a liberal decision) is
just .33,
49
but it increases to .46
50
when the public is very liberal. Figure
3 shows the predicted probabilities in between these extremes,
along with the 95% confidence intervals.
45 See THE SUPREME COURT DATABASE, supra note 22.
46 The 95% confidence interval is [.54, .63].
47 All other variables are set at their mean or mode.
48 The 95% confidence interval is [.66, .75].
49 The 95% confidence interval is [.30, .37].
50 The 95% confidence interval is [.41, .52].
278 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
FIGURE 4. PREDICTED PROBABILITY OF A LIBERAL DECISION IN A CIVIL
LIBERTIES CASE AS THE PUBLIC MOOD MOVES FROM VERY
CONSERVATIVE TO VERY LIBERAL
Note: This figure is based on the model in Table 2. The black lines indicate the
predicted probabilities in civil liberties cases and the gray lines shows 95% confidence
intervals. To generate these quantities of interest, we used S-Post. Civil Liberties is set at
1. All other variables are set at their means (Ideology of the Supreme Court, Ideology of the
President, Ideology of the House, Ideology of the Senate) or at their mode, which is 0 (U.S. Liberal
Party, U.S. Conservative Party, Economics, Judicial Power).
Whether the difference in the predicted probabilities displayed in
Figure 3 is large or small is a question readers must answer for themselves.
For what it’s worth, we think they are large—actually quite
large—given that we control for many other variables that have an
appreciable effect on the direction of the Court’s decisions.
The ideology of the Court, as represented in our model by the
ideal point of the median Justice, provides an example. Table 3 tells
us that the effect is statistically significant, such that the more conservative
the median, the less likely a liberal decision, and Figure 4
reveals that the effect is quite dramatic. All else being equal, the likelihood
of the Court reversing a conservative lower court decision
(i.e., reaching a liberal decision) when the median Justice is very lib-
Dec. 2010] WE’RE NOT SURE WHY 279
eral is a high .78.
51
The likelihood plummets to .55 when the median
is very conservative.
52
FIGURE 5. PREDICTED PROBABILITY OF A LIBERAL DECISION IN A CIVIL
LIBERTIES CASE AS THE MEDIAN JUSTICE MOVES FROM MOST
CONSERVATIVE TO MOST LIBERAL
Note: This figure is based on the model in Table 2. The black line indicates the predicted
probabilities in civil liberties cases and the gray line shows 95% confidence intervals. To
generate these quantities of interest, we used S-Post. Civil Liberties is set at 1 and the Direction of
the Lower Court Decision is conservative. All other variables are set at their means (Public Mood,
Ideology of the Supreme Court, Ideology of the President, Ideology of the House, Ideology of the Senate) or at
their mode, which is 0 (U.S. Liberal Party, U.S. Conservative Party, Economics, Judicial Power).
This finding is not especially surprising. On most realist theories
of judging—especially judging on the Supreme Court—the Justices’
ideology affects their decisions. What is surprising is that even after
taking into account ideology, Public Mood continues to be a statistically
significant and seemingly non-trivial predictor of outcomes (see
Table 3 and Figure 4).
51 The 95% confidence interval is [0.75, 0.81].
52 The 95% confidence interval is [0.53, 0.58].
280 JOURNAL OF CONSTITUTIONAL LAW [Vol. 13:2
IV. DISCUSSION
No doubt, our results should encourage Professor Friedman and
others who believe that “the Court and the public will come into basic
alliance with each other.”
53
At the least, our results indicate that an
association exists between the public’s mood and the Court’s deci-
sions.
Whether that association is something more—whether public
opinion “influences” the Court, as the subtitle of Professor Friedman’s
book asserts—we cannot say. This is so for two reasons. First,
despite our efforts to control for the more obvious determinants of
the Court’s decisions, our statistical model undoubtedly misses some
important variables. One indication of under-specification comes
from our own model. When we reestimated it, retaining the main effects
of Public Mood (Quarterly) and the issue area dummies but interacting
Mood with the issues, we found that public opinion was not a
good predictor of the outcome in Judicial Power cases. Because litigation
in this area tends to fly under the public’s radar screen, the
(non-)result implies the need to control for the importance of each
case.
54
Future research could take this step by deploying any number
of measures of salience, including the Epstein-Segal “New York Times”
approach,
55
the number of participating amici,
56
or even future citations
to the opinion.
57
A second reason we are unwilling to make the leap from association
to causality is we have neither posited nor tested a mechanism
53 FRIEDMAN, supra note 1, at 15.
54 But see Giles, Blackstone & Vining supra note 8, at 15 (finding that “the effects of public
opinion on liberalism appear to be consistently less in salient cases” because “case salience
may actually intensify the operation of [the Justices’ own ideological] prefe-
rences”).
55 Lee Epstein & Jeffrey A. Segal, Measuring Issue Salience, 44 AM. J. POL. SCI. 66, 72 (2000)
(measuring issue salience by whether the New York Times carried a front page story about
the case).
56 FORREST MALTZMAN, JAMES F. SPRIGGS II & PAUL J. WAHLBECK, CRAFTING LAW ON THE
SUPREME COURT: THE COLLEGIAL GAME 46 (2000) (measuring political salience by “determin[ing]
whether a case had more amici filings than the average case heard during a
term”).
57 See, e.g., James H. Fowler et al., Network Analysis and the Law: Measuring the Legal Importance
of Precedents at the U.S. Supreme Court, 15 POL. ANALYSIS 324, 326 (2007) (“It . . . seems reasonable
to determine how relevant a particular opinion is by considering how it is embedded
in the broader network of opinions compromising the law.”); Lee Epstein, William
M. Landes & Richard A. Posner, Why (and When) Judges Dissent: A Theoretical and
Empirical Analysis 14 (John M. Olin Law & Econ., Working Paper No. 510, 2d series,
2010), available at http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1542834 (“We
proxy importance by the number of Supreme Court and court of appeals citations to the
opinion.”).
Dec. 2010] WE’RE NOT SURE WHY 281
for the effect of public opinion on the Court. Certainly, Professor
Friedman and others could be right: the public has a direct influence
on the Court because the Justices are concerned about their legitimacy
in the short and long-terms. But it is equally plausible, as
Flemming and Wood explain, that the Justices are simply “social beings
confronted with the plethora of stimuli emanating from American
culture, media and politics.”
58
In other words, the same things
that influence public opinion may influence the Justices, who are, after
all, members of the public too.
Until we can behaviorally distinguish and assess these and other
mechanisms, statistical confirmation of Professor Friedman’s causal
story must wait yet another day. What we can say, and what our results
suggest, is that his account is not as implausible as some of the
political science literature suggests.
58 Flemming & Wood, supra note 10, at 471.