Law,Fact,and the Threat of Reversal From Above

Abstract

This article argues that the threat of review and reversal by supervising courts affects circuit court judges differently in disputes focusing on law compared to disputes focusing on facts. Because fact-bound cases are less likely to be reviewed than law-bound cases, lower court judges are freer to indulge their policy preferences in fact-bound cases. I test this argument using computer-assisted content analysis to measure the extent to which legal disputes are based on interpretations of facts and interpretations of relevant legal standards, respectively. The results of this content analysis are then used as independent variables in a model predicting the outcomes of legal challenges to the actions of administrative agencies. The results indicate that highly fact-bound decisions amplify the effects of judicial ideology while highly law-bound decisions constrain the effects of ideology.

Keywords

courts law content analysis strategic behavior

“If the facts are against you, argue the law. If the law is against you, argue the facts. If the law and the facts are against you, pound the table and yell like hell.”

—quoted by, Carl Sandburg, in The People, Yes (1936)

Carl Sandberg’s aphorism suggests that there are at least two possible bases for resolving legal disputes, and that these different bases can point toward different outcomes. Litigants can present arguments based on interpretations of relevant legal standards (i.e., law) or on interpretations of the relevant facts, or both, and judges can explain their resolution of the dispute on either of these bases. It may be, however, that one of these bases constrains the judges more than the other, and that the fact-bound or law-bound nature of the case influences how judges decide these cases.

The question of whether and to what extent law constrains judges has been at the center of scholarly debates in the judicial politics field for the last five decades and has animated countless research projects. The question is important because the policymaking role of judges in our political system only makes sense if judges’ decisions are guided by law. A great deal of scholarly research has demonstrated that policy preferences play a significant role in judicial decisions. This effect is strongest at the Supreme Court level (Segal & Spaeth, 1993) but is also significant at the Federal Courts of Appeals (Goldman, 1966, 1975; Sunstein, Sawicki, Schkade, & Ellma, 2006) and, less so, at the Federal District Courts (e.g., Songer & Johnson, 2002). This research has pushed scholars of judicial politics to view the role of law skeptically. An important question is the circumstances under which the law is more or less influential in the outcomes of court decisions.

I advance this area of study by exploring whether the nature of a legal dispute, the extent to which it focuses on disagreement over facts or disagreement over law, systematically increases or decreases the influence of judges’ policy preferences. I find that the more a dispute focuses on facts, the more influence the judges’ policy preferences have on the outcome of the case. Conversely, the more a dispute centers on the proper meaning and application of legal standards, the less influence judges’ policy preferences have on the outcome of the case. In disputes focused almost exclusively on law, liberal judges and conservative judges act very much alike. The likelihood of review by higher courts is the most plausible explanation for this difference.

In order to determine the bases of the courts’ decisions, I dig into judicial opinions. Judicial scholars have long acknowledged that the content of judicial opinions is important,¹ but have rarely tackled the task of incorporating opinion content into their models (but see Corley, Howard, & Nixon, 2005, Owens & Wedeking, 2011).² Computer-assisted content analysis (CACA) offers a reliable and economical tool for measuring the content of opinions. CACA overcomes the problems of inconsistency in human coding of texts and has the benefit of transparency because it allows the researcher to specify exactly what features of the texts are utilized. This method has great potential for improving scholars’ ability to study the content of legal texts in systematic ways.

This research advances current knowledge of judicial behavior in two important ways. First, it shows that the fact-bound or law-bound nature of court cases systematically affects the extent to which judges rule according to their policy preferences. The more a case focuses on factual matters, the more magnified is the effect of the judges’ preferences. Increased focus on legal standards dampens the effect of policy preferences. When cases turn on factual matters, variation in judges’ policy preferences generates wide variation in outcomes. When cases turn on the proper interpretation of legal standards, judges with very different policy preferences produce very similar outcomes. I interpret this finding as showing that legal standards constrain judges’ decisions because disputes based on legal standards are more likely to trigger review and possible reversal by higher courts. Law is less influential in fact-bound cases because judges know these cases are very unlikely to be reviewed.

Secondly, this article describes and demonstrates a method of CACA that can be used to characterize legal texts (e.g., court opinions, briefs submitted by litigants or amici, or statutes) in terms of their content. This method allows researchers to include the contents of legal documents in their models in a way that is economical and meets social science standards of validity and reliability. The method is also very flexible, in that it can be adapted to measure any dimension of content.

The article proceeds as follows: the next section explains the theoretical assumptions that support the argument and spells out the relationships between the fact-bound or law-bound nature of legal disputes and the influence of judges’ policy preferences. The third section explains the research design in detail. The final sections of the article present the results of the analyses and discuss their implications.

Legal and Strategic Effects on Circuit Court Decisions

Research into decision making on the U.S. Courts of Appeals shows that decisions are guided by the policy preferences of the judges (e.g., Baum, 1997; Goldman, 1975; Humphries & Songer, 1999; Songer & Davis, 1990; Sunstein et al., 2006), relevant legal standards (Cross, 2007; Songer, Ginn, & Sarver, 2003), the preferences of supervising courts (Haire, Songer, & Lindquist, 2003; Songer, Segal, & Cameron, 1994) and the particular combinations of judges on judicial panels (Kastellec, 2011). Taken as a whole, this research suggests that while judicial policy preferences are important, they are not the only systematic influence.

Scholars have refined this research by looking for circumstances in which judges’ preferences are more or less influential. Recent studies have identified several legal factors that moderate the influence of judges’ policy preferences on their decisions. At the Supreme Court level, Bartels (2009) demonstrates that the influence of justices’ ideologies on their decisions varies in predictable ways. Zorn and Bowie (2010) show that the place of a court within the judicial hierarchy moderates the effect of policy preferences, with preferences mattering more as cases move up the judicial ladder. Lindquist and Cross (2005), studying decisions in the U.S. Courts of Appeals, show that judges’ policy preferences are significantly more influential in cases of first impression (for which no authoritative precedent exists) than in other cases. Similarly, Randazzo, Waterman, and Fine (2006) demonstrate that detailed statutes constrain the extent to which judges indulge their policy preferences when making decisions. All of these studies identify elements of the contexts of legal disputes that structure the effect of judges’ policy preferences on case outcomes.

A separate vein of research has looked for strategic behavior by circuit court judges. The strategic theory of judging asserts that judges try to achieve their policy goals to the maximum extent possible while minimizing the policy and personal costs of reversal. A couple of studies have tested the proposition that Federal Circuit Court judges look ahead to the possibility of review by higher courts and adjust their policymaking actions accordingly. Smith and Tiller (2002) showed that panels of circuit court judges strategically choose the legal bases for their decisions in order to manage the probability of review by higher courts. They found that judges were more likely to choose legal bases that were resistant to review by higher courts for decisions whose policy outcomes the judges favored. Hume (2009) reported similar findings. These studies indicate that judges do anticipate the reactions of future policymakers to their decisions and take steps to protect decisions they favor on policy grounds.

Other studies, however, cast doubt on the notion that circuit court judges take steps to avoid reversal from above. Klein and Hume (2003) found that the likelihood that a particular dispute would be reviewed by the Supreme Court did not affect whether the circuit court issued a decision consistent with the Supreme Court’s likely preferences. Klein and Hume, however, did not measure whether the effect of the circuit court judges’ ideologies was influenced by the likelihood of review. More recently, Bowie and Songer (2009) argued that anticipation of possible review by the Supreme Court plays virtually no role in circuit court decisions because judges don’t particularly mind being reversed and because it is very difficult to predict which cases will be reviewed. These mixed findings indicate that the jury is still out on whether circuit court judges do act strategically in this way. The evidence I present here supports the strategic perspective on judging.

Law, Fact, and Review by Higher Courts

Legal disputes can focus on factual matters or on legal standards, or a mixture of the two.³ Disputes focusing on facts involve questions about the accuracy and relevance of the facts asserted by the litigants and whether or not the facts submitted by the instigating party are sufficient to meet the appropriate burden of proof or standard of review. When an appellate court is reviewing the decision of a trial court, the appellate court will typically accept the factual conclusions of the trial court unless it finds that they are “clearly erroneous.” In the area of administrative law, disputes over facts frequently turn on whether the agency successfully shows that its action is supported by “substantial evidence.”⁴ These criteria are deferential, but not precise. There is no obvious rule for determining whether a decision being reviewed is supported by enough accurate and relevant evidence for it to be termed “substantial” or not “clearly erroneous.”

Resolving factual disputes is not merely a matter of determining what the true facts are. Often there are undisputed facts supporting both sides. The court must decide which facts offer the most relevant and compelling evidence. Neely v. Shalala (1993) ⁵ provides an example of a dispute focused on facts. Neely applied for social security disability benefits, which were denied. The court’s task was to determine whether the Social Security Administration’s denial of Neely’s application was supported by substantial evidence. The case turned on whether Neely was sufficiently disabled by pain to be unable to work. Resolving the dispute involved not only weighing Neely’s testimony with regard to her pain against evidence that she was able to carry out household tasks and even go deer hunting without pain medication, but also determining whether jobs were available that Neely would be able to perform. The court’s opinion treats the legal rule as settled and focuses on determining how the facts relate to the legal rule.⁶ The reasoning in this type of opinion is tied to the specific facts of the case and does not generate broad rules applicable to other disputes.

State Corporation Commission v. Interstate Commerce Commission (ICC), (1989)⁷ provides a different example of a dispute centered on facts. In this case, a Tenth Circuit Court Panel evaluated the ICC’s decision that Missouri Pacific’s abandonment of 66 miles of railroad line was economically justified and would not cause serious problems for the community served by the line. As the court noted, the State Corporation Commission was “primarily challenging several findings of the ICC and the sufficiency of the underlying evidence” (894 F.2d 1142). The evidence in the case concerned how much money Missouri Pacific would save by abandoning the line, whether alternative modes of transportation could provide for the needs of the local area, and the increased road maintenance costs that would result from trucking of cargo that had previously moved by rail. Finding that Missouri Pacific presented “plausible” reasons for abandoning the rail line, the court upheld the ICC’s decision. The circuit panel had to make relatively unstructured evaluations of the credibility and evidence offered by the two sides, and then apply the vague “substantial evidence” standard. Fact-based disputes like this one are less amenable to articulating legal policy than law-based disputes. Consequently, the Supreme Court is unlikely to review this type of case (Perry, 1991).

In contrast to disputes focusing on facts, disputes over legal standards require judges to identify the legal standard relevant to the dispute and derive from it a decision rule for resolving the current dispute. For a number of reasons, this is a more highly structured task than evaluating factual matters. The raw materials of this type of analysis (statutes, regulations, previous court decisions, and legislative history) are both limited and familiar to all judges. This makes it more likely that higher courts will be confident they understand all aspects of the dispute and therefore more assertive in reviewing them. More importantly, these types of disputes create standards that will be applicable to other cases and are therefore good vehicles for articulating legal policy.

A 1988 case, Brotherhood of Railway Carmen v. ICC (1988) ,⁸ illustrates the task of resolving a dispute over the meaning of a legal standard. In this case, a D.C. Circuit Court panel had to determine whether section 11341(a) of the Interstate Commerce Act⁹ allowed the ICC to override a collective bargaining agreement between unions and the railroad carrier CSX. The panel’s opinion examines whether statutory language allowing the ICC to exempt carriers from “antitrust laws and from all other law” authorizes the ICC action. The opinion discusses the statutory language, relevant Supreme Court precedent, and the legislative history of the Act on the way to concluding that the ICC exceeded its statutory authority. This sort of dispute generates an interpretation of a statute that can be applied to other, similar disputes and thereby establishes legal policy. If the supervising court disagrees with that legal policy, it is likely to review the case and replace the lower court’s policy with its own preferred policy.

Beyond these theoretical arguments, there is significant empirical evidence that the Supreme Court is unlikely to review cases focused on facts and evidence. Perry (1993, pp. 223-224) reports that disputes over the sufficiency of evidence are particularly unlikely to make it through the Supreme Court’s certiorari process. Bowie and Songer (2009, p. 401) find that circuit court cases focused on “weight of evidence” or the substantial evidence standard are 61% and 85% less likely to be reviewed by the Supreme Court, all other factors being equal. Although it may be difficult for circuit judges to predict which cases will be reviewed by the Supreme Court, it is not difficult to predict that fact-bound cases are very unlikely to be reviewed.

The ability to predict which cases will escape the higher court’s scrutiny is useful. A confident expectation that fact-bound cases will not be reviewed would plausibly lead circuit judges to exercise their policy preferences more in these sorts of cases than in disputes over legal standards. Further, judges’ expectations that fact-bound cases will not be reviewed and that higher courts will look at the opinion resolving the dispute to determine whether the cases focuses on facts or law could lead them to manipulate the content of the opinion. When judges want to protect a decision that matches their policy preferences, they would package the decision in an opinion that stresses facts rather than law. This is consistent with the conclusions of Smith and Tiller (2002) and Hume (2009). Disputes that can be presented as fact-bound invite greater exercise of judges’ policy preferences because such disputes are very unlikely to be reviewed by higher courts, while the higher probability of review associated with disputes focused on legal standards dampen the effects of judges’ policy preferences. These factors suggest two hypotheses:

Hypothesis 1 (H1): The influence of judges’ policy preferences on case outcomes will increase as opinions focus more on arguments over facts.

Hypothesis 2 (H2): The influence of judges’ policy preferences on case outcomes will decrease as opinions focus more on interpretations of legal standards.

Research Design

My strategy for evaluating these hypotheses involves creating indexes of the extent to which court cases involve disputes over facts and the extent to which they involve disputes over legal standards. I then use these indexes as independent variables and estimate their effects on the extent to which policy preferences influence the court’s decision. The data consist of administrative law cases drawn from United States Courts of Appeals Database.¹⁰ I selected all decisions involving disputes that originated in federal administrative agencies from the years 1925-2002.¹¹ The data end in 2002 because that is the most recent year contained in the Courts of Appeals Database. Administrative law furnishes a good context for this research because administrative law cases separate conceptually into disputes over the factual basis of the agency’s decision and disputes over the agency’s legal authority.¹² The dependent variable in these analyses is a dummy variable coded 1 if the agency wins the case, and coded 0 if the agency loses the case in whole or in part. The variable is coded 1 in about 53% of the cases.

Main Explanatory Variables

The main independent variables of interest are the indexes that indicate the extent to which each case focuses on factual matters and legal standards, respectively. I generate these variables through content analysis. Systematically coding the content of a large set of opinions is difficult and time-consuming. Computer-assisted coding has advantages over human coding in terms of time, reliability, and transparency (Evans, McIntosh, Lin, & Cates, 2007, Hall & Wright, 2008).¹³ These qualities have led political scientists to adopt CACA for coding party platforms (Laver, Benoit, & Garry, 2003), presidential speeches (Coe, 2007), and the Congressional Record (Monroe, Colaresi, & Quinn, 2008).

Legal documents are well-suited for CACA because they are highly structured in terms of the vocabulary used. Words and phrases have meanings that are understood by the community of lawyers and judges who write and read them. This means that participants in the legal process who want to convey a particular message must use a relatively limited and predictable set of terms (Gibbons, 2003). In the study of judicial politics, Corley and coauthors (Corley, 2008; Corley, Collins, & Calvin, 2011) have made innovative use of plagiarism software to measure the extent to which the Supreme Court takes opinion language from litigants briefs and lower courts. McGuire and Vanberg (2005) and McGuire, Vanberg, and Yanus (2007) have applied the Wordscores software (Laver et al., 2003) to characterize the ideological tone of Supreme Court opinions. The type of CACA I use here is more flexible than the approaches discussed above, and is more similar Evans et al.’s (2007) classification of amicus briefs according to their ideological direction.

The goal of the content analysis is to produce indicators of the extent to which the opinion in each case focuses on factual matters and legal standards, respectively. These categories are not mutually exclusive; some opinions discuss both types of questions extensively. Therefore, I develop two separate indexes for each opinion, one indicating the extent to which the opinion discusses factual matters and one indicating the extent to which the opinion discusses legal standards. These indexes are sets of terms (i.e., words and phrases) that are associated with each type of analysis, factual and legal review. Each index is the cumulative frequency of terms that are highly indicative of the type of review.

The first step in creating the indexes was to create a list of candidate terms that might be associated with disputes over facts, or law, respectively. I assembled this candidate list by creating a frequency count of all the words used in all the opinions. I then examined this list for terms that were used relatively frequently and seemed, from my knowledge of administrative law, likely to be used in discussion of factual matters and legal standards, respectively. To complete the list, I added several phrases that are closely associated with the one of these types of legal arguments.¹⁴ Although the creation of this list of candidate terms involves subjective elements, the list itself is only a starting point and includes any terms that could plausibly indicate a discussion of legal standards or factual bases relevant to administrative law disputes. The complete list of candidate terms is presented in Appendix A.

I then employed a computer program which examined each court opinion and recorded the frequency of each candidate term in each opinion.¹⁵ I transformed these raw frequencies in the following manner to make them more comparable across opinions and across terms: I divided each term’s frequency by the length of the opinion in words to get the relative frequency of the term¹⁶ and then divided these relative frequencies by the mean relative frequencies across all opinions.¹⁷ This transformation produced standardized frequencies that reflect the frequency of each term, relative to its average frequency. So, a standardized frequency higher than 1 indicates the term was used with greater than its average frequency in that opinion; a standardized frequency lower than 1 indicates the terms was used with lower than its average frequency in that opinion. These frequency data were then merged with the United States Courts of Appeals Database. This produced a data set of around 1,500 cases with information on the standardized frequencies of the candidate terms in each opinion and all the information on the origins, participants, issues, and outcomes of the cases contained in the Courts of Appeals Database.

To determine which terms were associated with factual matters and legal standards, respectively, I manually coded a training set of 142 opinions as to whether the opinion discussed factual matters, legal standards, or both.¹⁸ Of these 142 cases, 82 involved legal standards and 89 involved factual matters.¹⁹ A list of the citations of these cases and how each was coded is included in Appendix B.

I used these classifications of opinions to determine which words and phrases are associated with legal standards and factual matters, respectively. For each term in the list of candidate terms I calculated the mean standardized frequency value for the manually coded opinions involving legal standards and the mean standardized frequency value for the manually coded cases not involving legal standards and take the difference between these means. This difference is large for terms which are central to the content of opinions which resolve a dispute over legal standards but not relevant to the content of opinions not involving legal standards. Therefore, the size of this difference indicates the value of the term for distinguishing between opinions that involve legal standards and those that do not. I followed the same procedure to determine the differences in mean standardized frequency values between opinions involving factual matters and those not involving factual matters for all terms in the candidate list.²⁰ After calculating these differences of means, I created the two indexes by selecting all the terms for which the difference of means is positive and significant at the .01 level. Twenty-three terms met this test for opinions involving legal standards; 15 terms satisfied the test for opinions involving factual matters. The terms in each index are listed in Table 1.²¹

Table 1.

Terms in Each of the Language Indexes.

Terms associated with legal review	Terms associated with factual review
Application	Conclusion, conclusions
Authority	Credible, credibility
CFR (Code of Federal Regulations)	Evidence, evidentiary
Chapter	Fact, facts, factual
Congress	Finding, findings
Construction, constructions	Found
Construe, construes, construed	Proof, prove, proves, proved
Defer, deference, deferential	Record
Intent, intended, intention	Statement
Jurisdiction	Substantial, substantially
Language, languages	Support, supports, supporting
Legal, legally	“Substantial evidence”
Legislation, legislative, legislature	Test
Meaning, meanings	Testify, testified, testimony
Process	“Trial Examiner”
Provision, provisions
Regulation, regulations
Require, requires, required, requirements
Review, reviewed, reviews
Statute, statutes, statutory
Text, texts, textual
Unreasonable, unreasonably

The terms in Table 1 have strong face validity. From the first column, terms such as chapter, Congress, CFR, jurisdiction, language, legal (and variants), legislation (and variants), provisions, regulations (and variants), statute (and variants), and text (and variants) all refer to law or sources of law. Application, defer (and variants), intent (and variants), meaning, and review (and variants) all refer to the process of interpreting legal standards, and unreasonable or unreasonably are often used to describe the way a law has been interpreted or applied. Authority and require (and variants) relate to the way statutes control behavior.

Likewise, the terms in the second column clearly relate to factual review. Record and statement refer to the administrative record that contains the facts relevant to the dispute. Terms such as credible, evidence, facts, findings, proof, showing, support, and testimony (and variants of all these) all relate to the evidence relevant to the dispute. Trial Examiner refers to judges who determined the facts of the case before the dispute arrived at the Circuit Court. Substantial evidence is the usual standard applied in fact-based disputes in administrative law.

The lists of words in the indexes were transformed into numeric values by multiplying each term by the standardized frequency of the term in the opinion. These indexes indicate the extent to which each dispute concerns factual matters and legal standards, respectively.

To create the interactive variables used to evaluate my hypotheses, I multiplied these indexes by a measure of the circuit court judges’ ideological sympathy for the agency decision being challenged. Panel Alignment with Agency measures the extent to which the policy preferences of judges on the circuit court panel are consistent with the agency action. Circuit court decisions are ordinarily made by panels of three judges. I used the median Giles-Hettinger-Peppers (GHP) score on the panel to represent the policy preferences of the judges (Giles, Hettinger, & Peppers, 2001). These scores range from –.684 to .581, with higher scores indicating more conservative preferences. The direction of the agency action being reviewed is derived from the Courts of Appeals data set. If the agency action is conservative, Panel Alignment with Agency takes on the median GHP score on the panel. If the agency action is liberal, Panel Alignment with Agency is the median GHP score multiplied by −1. Higher values of Panel Alignment with Agency indicate greater panel agreement with the agency policy being challenged. In the analyses presented below, the Panel Alignment with Agency variable represents the influence of the judges’ policy preferences on the case outcome.

The interactive variables Panel Alignment*Legal Language and Panel Alignment*Factual Language are the multiplicative products of their components. The arguments presented above suggest that the influence of Panel Alignment with Agency should increase as the Index of Factual Language increases, and therefore the coefficient on this interactive variable should be positive. A second implication of the theoretical argument is that the influence of Panel Alignment with Agency should decrease as the Index of Legal Language increases, and therefore the coefficient on this interactive variable should be negative.

Control Variables

Beyond these variables of primary interest, I include several variables to control for other plausible influences on the courts’ decisions. Agency Alignment with Supreme Court measures the ideological consistency between the agency action being reviewed and the policy preferences of the Supreme Court. It is the median Judicial Common Space (JCS) score for the U.S. Supreme Court (Epstein, Martin, Segal, & Westerland, 2007).²² If the agency action is conservative, Agency Alignment with Supreme Court takes on the median Judicial Common Space (JCS) score on the Court. If the agency action is liberal, the Agency Alignment with Supreme Court is the median JCS score multiplied by −1. In this way, higher numbers represent a greater probability that the Supreme Court would approve of the agency action. If the circuit court panels are attempting to implement the preferences of the Supreme Court, this variable should be positively associated with decisions favoring the agency. Agency Alignment with Circuit measures the ideological consistency between the agency action being reviewed and the median GHP score of the judges on the circuit in which the case is being heard. This variable is coded similarly to Agency Alignment with Supreme Court. If circuit court panels are influenced by the preferences of their circuit, this variable should reflect that influence.

Agency is Appellant is a dummy variable coded 1 when the agency is the party appealing the verdict below. If the agency is the appellant, this means that it was not the winner in the court below. This indicates that it may not have a strong basis for its action, and so this variable is expected to be negatively associated with decisions favoring the agency. DC Circuit is a dummy variable indicating that the panel hearing the case is part of the DC Circuit Court of Appeals. The DC Circuit’s administrative law jurisdiction includes most regulations that have national effect. These regulations are more complicated to defend than narrower rules or adjudications, so the agencies’ winning percentage in the DC circuit is likely to be lower than in other circuits. If this is the case, DC Circuit should be negatively related to Agency Win.²³ Summary statistics for all variables are presented in Appendix C

I use logit regressions to estimate the effects of the independent variables on the probability that the court will rule in favor of the agency.²⁴

Results

The results from logit regressions, presented in Table 2, support the hypotheses that the effects of judges’ policy preferences depend on the factual or legal nature of the dispute. The first column of numbers evaluates the effects of factual language, and the second column of numbers relates to the effects of legal language. Both models show highly significant chi-square statistics, and both show significant predictive ability, reducing the error rates by at least 23% each.

Table 2.

Results for Logit Analyses of Influences on Agency Win.

	Effect of fact-related language	Effect of law-related language	Both law and fact-related language
Panel alignment with agency	.371*	.313	.407**
Panel alignment with agency	(.153)	(.158)	(.147)
Index of factual language (mean-centered)	−.005		−.003
Index of factual language (mean-centered)	(.004)		(.004)
Panel alignmentfactual language*	.051***		.044***
Panel alignmentfactual language*	(.014)		(.011)
Index of legal language (mean-centered)		.007	.007
Index of legal language (mean-centered)	(.004)		(.004)
Panel alignmentlegal language*		−.023***	−.013*
Panel alignmentlegal language*		(.007)	(.006)
Agency is appellant	−0.704**	−.735**	−0.730**
Agency is appellant	(0.256)	(.258)	(0.261)
Agency alignment with supreme court	.048	.065	.063
Agency alignment with supreme court	(.315)	(.338)	(.320)
Agency alignment with circuit	.575**	.578**	.569**
Agency alignment with circuit	(.185)	(.190)	(.182)
DC circuit	−.228**	−.224	−.243
DC circuit	(.074)	(.071)	(.146)
Constant	.588***	.546***	.536***
Constant	(.149)	(.161)	(.159)
N	1,497	1,497	1,497
Wald χ²(df)	718.8 (11)***	558.3 (11)***	429.9(11)***
PRE	25.6%	23.0%	24.3%

Note. Fixed effects for agencies were included in analysis but are not reported.

Effect of Factual Language

Looking first at the effects of fact-related language, the coefficient for Panel Alignment with Agency, .371, is positive and statistically significant. Because of the interactive variable in the model, this coefficient reflects the situation when Index of Factual Language equals it mean.²⁵ The fact that the coefficient on the interaction term, Panel Alignment*Factual Language, is positive and much larger than its standard error suggests that the effect of Panel Alignment increases as the amount of factual language increases. However, a better understanding of these results can be gained by looking at Figure 1.²⁶

Figure 1.

Effect of judges’ ideologies as factual language increases.

The solid line trending upward from left to right in Figure 1 reflects the influence of the judges’ policy preferences as the amount of fact-related language in the opinion increases. The two dashed lines above and below the solid line represent the 95% confidence interval for this influence. The vertical line bisecting the graph indicates the point at which the 95% confidence interval rises above 0. On the right side of the graph, the sloping line and the lower 95% confidence interval are both above 0, indicating that in this region the effect of the judges’ policy preferences is positive and significant. This is consistent with the attitudinal model of judicial voting, which suggests that policy preferences are a major influence on judicial votes.

On the left side of the graph, however, the lower 95% confidence interval is below 0, indicating that the effect of the judges’ policy preferences is not significant at low values of the Index of Factual Language. The lower boundary of the confidence interval crosses 0 when the Index of Factual Language equals approximately 13.2, which is just above the median point (56th percentile) of the distribution of factual language index scores. Figure 1 illustrates that the influence of judges’ policy preferences increases as disputes focus more on fact-related review, and that for disputes below the 56th percentile in terms of factual content, judges’ policy preferences do not exert a statistically significant influence on their decisions.

Effect of Legal Language

The second column of numbers in Table 2 pertains to the relationship between the amount of legal language in the opinion and the judges’ policy preferences. The second column of numbers in Table 2 pertains to the influence of legal language on the effect of judges’ policy preferences. Again the coefficient for the Panel Alignment with Agency is positive and statistically significant, indicating that when the Index of Legal Language is at its mean, the judges’ policy preferences are shown to influence case outcomes. The coefficient on the interaction term, Panel Alignment*Legal Language, is negative and statistically significant at the .001 level, suggesting that the effect of Panel Alignment decreases as the amount of legal language increases. Again, this is consistent with theoretical argument above.

Figure 2 shows the changing effect of Panel Alignment with Agency as the amount of legal language in the opinion increases. Again, the dashed lines indicate the 95% confidence interval of the effect. As factual language increases, judges’ policy preferences become less important to the case outcome. Panel Alignment with Agency does not significantly influence case outcomes once Index of Legal Language rises above 22.5, which is about the 60th percentile of the index. Thus, for disputes that are above the 60th percentile in terms of their focus on statutory matters, judges’ policy preferences are not statistically significant.

Figure 2.

Effect of judges’ ideologies as legal language increases.

The third column of numbers in Table 2 presents the results of a model including both the statutory language and factual language indexes and each of these variables interacted with Panel Alignment with Agency. The coefficient for Panel Alignment * Factual Language is again positive and statistically significant (p < .001), indicating that judges’ attitudes become more important as the disputes focus more on factual maters. The coefficient for Panel Alignment*Legal Language is again negative and statistically significant (p < .05). These consistent results strongly indicate that appellate judges’ ideologies are more influential in fact-bound cases and less influential in law-bound cases.

Changes in Probability of Agency Win

Another way to understand how the fact-bound or law-bound nature of a dispute influences the outcome of the case is to analyze changes in the probability of a particular case outcome. Figures 3 and 4 present predicted probabilities of circuit court panels upholding agency decisions.²⁷ Figure 3 presents the effects of changes in the Index of Factual Language. The dashed line sloping upward from left to right shows the effect of an increasing amount of factual language on the predicted probability of a ruling in favor of the agency when the judges are ideologically sympathetic to the agency action (i.e., a liberal panel of judges reviewing a liberal agency action or a conservative panel reviewing a conservative agency action).²⁸ The dark shaded area represents the 95% confidence interval around the predicted probabilities. The solid line sloping downward from left to right shows the effect of an increasing amount of factual language on the probability of a ruling in favor of the agency when the judges are ideologically hostile to the agency action. The lighter shaded area around this line shows the 95% confidence interval. Note that for cases with a relatively small amount of factual language (to the left of the vertical line), the confidence intervals for the two probability lines overlap. The confidence intervals are separate only for the area of the graph on the right side of the vertical line, which is at the 62nd percentile of the Index of Factual Language. Thus, for cases in the bottom 62% in terms of factual language content, there is no significant difference between liberal and conservative judges. For highly fact-bound cases, an ideologically sympathetic panel is 33% more likely to affirm the agency than an ideologically hostile panel.

Figure 3.

Effect of ideology and factual language on case outcomes.

Figure 4.

Effect of ideology and legal language on case outcomes.

Figure 4 presents similar information on the effect of legal language. As in the previous figure, the dashed line indicates the probability that an agency decision will be upheld by a panel that is ideologically sympathetic to the decision, and the dark shaded area shows the 95% confidence interval of this probability. The solid line and lighter shaded area pertain to a judicial panel that is ideologically hostile to the agency policy. On the left side of the figure, the two confidence intervals are separate, indicating that the ideology of the panels makes a difference. At the extreme left side of the figure, sympathetic panels are 16% more likely to affirm the agency decision than hostile panels. As cases focus more on legal standards, the two confidence intervals overlap. The vertical line, which is at the 52nd percentile of the Index of Legal Language, shows the point at which the confidence intervals separate. For cases in the top 48% in terms of legal language content, there is no significant difference between liberal and conservative judges.

Conclusion and Implications

The results presented here suggest an addendum to Sandberg’s aphorism: “If the judge is with you, argue the facts. If the judge is against you, argue the law.” The effect of judges’ policy preferences is magnified in disputes that focus on factual matters, and dampened in disputes that focus on legal standards. The most plausible explanation for these results is that law constrains judges indirectly: judges believe cases decided on facts are less likely to be reviewed by higher courts, and therefore see in factual cases opportunities to indulge their policy goals with little fear of reversal.²⁹ A slightly different explanation would be that judges may be emphasizing the factual elements of disputes whose outcomes match their policy goals and manipulating the contents of these opinions to make the cases look like they were about facts. This is essentially what Smith and Tiller (2002) and Hume (2009) assert is happening.

One worthwhile extension of this research would be to investigate which of these two processes are driving the relationship between the nature of the court opinions and the influence of judicial ideology. This could be investigated by using content analysis to compare the kinds of arguments made in litigants’ briefs with the arguments discussed in the resulting court opinions.³⁰ If judges show a pattern of writing fact-based opinions in cases that were argued on the basis of legal standards when ruling in favor of their own policy preferences, but not when ruling against their policy preferences, this would be very strong evidence in support of the strategic model of opinion-writing.

It is not clear how well these findings would translate to other courts. A key factor is the discretionary docket of the circuits’ supervising courts. That is, because the U.S. Supreme Court and the circuits en banc can choose which cases to review, three-judge circuit court panels have a reason to make some decisions more or less inviting to review. This would not be true of the Federal District Courts or of the U.S. Supreme Court. However, many state supreme courts do have discretionary docket control (Langer, 2002), and so the same phenomena documented here might well be evident among state courts one level below the state Supreme Court.

This research has broader implications for the role of law in guiding judges’ decisions. Because law-bound disputes have the potential to generate broadly applicable legal rules, judges may be relatively more motivated by the implications of the decision rule articulated in the opinion compared to fact-bound disputes. In fact-bound disputes, it may be that the judges’ preferences would be focused more on their attitudes toward the parties to the suit, whereas in law-bound disputes the judges’ preferences would be more focused on the content of the opinion.

Footnotes

Appendix A.

List of all Candidate Terms and Phrases.

Terms and phrases expected to be associated with legal review	Terms and phrases expected to be Associated with Factual Review
• 706(2)(b)	• 706(2)(a)
• 706(2)(c)	• 706(2)(e)
• 706(2)(d)	• 706(2)(f)
• Accordance, according	• Adequate, adequately
• Adjudication, adjudications	• Arbitrary
• Apply, applied	• Average
• Application	• Basis, bases
• Arbitrary	• Capricious
• Authority	• Concluded
• Authorized, authorizes, authorization, authorize	• Conclusion, conclusions
• Canon	• Data
• Chapter	• Demonstrate, demonstrates
• Clause	• Determination
• Code	• Determine, determined, determining
• Congress	• Discretion, discretionary
• Construction, constructions	• effect
• Construe, construes, construed	• Error
• Contrary	• Estimate, estimation, estimates, estimated
• Controlling	• Evidence, evidentiary
• Cost, costs	• Expert, experts, expertise
• Defer, deference, deferential	• Explain, explains, explanation
• Define, defines, defined, definition	• Fact, facts, factual
• Delegate, delegates, delegated, delegation	• Factors, factors
• Dictionary	• feasible
• Document	• Find, finds, findings
• duty	• Information
• Illegal, illegally	• Insubstantial
• Intent, intention, intended	• Insufficient
• Interpret, interprets, interpreted, interpretation, interpretations	• Insufficiently
• Jurisdiction	• Irrational, irrationally
• Language, language’s	• Logic, logically, logical
• Law, laws, lawful, lawfully, lawfulness	• Measure, measured, measures
• Unlawful, unlawfully, unlawfulness	• Median
• Legal, legally	• Method, methods, methodological, methodologies
• Legislation, legislature, legislative	• Percent
• Meaning, meanings	• Likelihood
• Nondiscretionary	• Probability, probabilistic
• Notice	• Properties
• Order	• Proof, prove, prove, proved
• Orders	• Quantitative, quantitatively, quantify, quantification, quantitative
• Permissible	• Quality, qualitative, qualification, qualify
• Precedent, precedents	• Rational
• Prior	• Rationale
• Previous	• Reason, reasoned, reasons, reasoning
• Provision, provisions	• Reasonable, Reasonably
• Procedure, procedures, procedural	• Record
• Process	• Relevant, relevance
• Proper	• Risk
• Properly	• Science
• Purpose	• Scientific
• Pursuant	• Showing
• Read, reads, readings	• Statement
• Regulation, regulations	• Statistic, statistics, statistical
• Reinterpret, reinterpreted	• Study, studies, studied
• Require, requires, required, requirement	• Substantial, substantially
• Right, rights	• Support, supports, supporting
• Rule, rules	• Sufficient, sufficiently
• Rulemaking	• Technical
• Scope	• Testify, testified, testimony
• Section	• Test
• Silent	• Uncertainty
• Statute, statutes, statutory	• Arbitrary and capricious
• Term, terms	• Substantial evidence
• Text, texts, textual	• Rational basis
• Ultra	• Reasoned decision
• Unreasonable, unreasonably	• Reasonable basis
• Vires	• Basis and purpose
• Vocabulary	• Unwarranted by the facts
• Word, words	•
• In law	•
• Of law	•
• unlawfully withheld	•
• Unreasonably delayed	•
• Abuse of discretion	•
• Accordance with law	•
• Statutory jurisdiction	•
• Statutory right	•
• Procedures required by law	•
• Ultra vires	•

Appendix B.

List of Cases Hand-Coded and Categorization as Involving Legal Review and Factual Review, Respectively.

Case citation	Factual review	Statutory review
113 F.2d 667	Yes	No
115 F.2d 681	No	Yes
119 F.2d 131	Yes	No
119 F.2d 561	Yes	Yes
123 F.2d 90	Yes	Yes
138 F.2d 884	Yes	No
190 F.2d 576	Yes	No
286 F.2d 158	Yes	No
287 F.2d 469	No	Yes
288 F.2d 818	Yes	No
292 F.2d 770	No	Yes
305 F.2d 763	Yes	Yes
308 F.2d 230	Yes	No
310 F.2d 89	Yes	No
311 F.2d 1	Yes	No
317 F.2d 912	Yes	No
325 F.2d 126	Yes	No
326 F.2d 488	Yes	No
331 F.2d 165	Yes	No
336 F.2d 115	Yes	No
336 F.2d 942	Yes	Yes
344 F.2d 47	No	Yes
351 F.2d 771	Yes	Yes
355 F.2d 851	Yes	No
360 F.2d 856	No	Yes
361 F.2d 300	Yes	No
365 F.2d 515	Yes	No
369 F.2d 495	Yes	Yes
375 F.2d 707	Yes	No
376 F.2d 131	Yes	No
379 F.2d 536	Yes	No
379 F.2d 814	Yes	No
383 F.2d 466	Yes	Yes
385 F.2d 981	No	Yes
390 F.2d 304	Yes	No
391 F.2d 713	Yes	No
391 F.2d 961	Yes	No
394 F.2d 84	No	Yes
395 F.2d 622	Yes	No
397 F.2d 801	Yes	No
404 F.2d 1370	Yes	No
406 F.2d 1306	Yes	No
412 F.2d 37	Yes	Yes
415 F.2d 78	Yes	No
416 F.2d 243	Yes	No
447 F.2d 290	Yes	Yes
463 F.2d 256	No	Yes
469 F.2d 498	Yes	No
489 F.2d 1247	Yes	Yes
500 F.2d 597	Yes	Yes
501 F.2d 191	Yes	Yes
505 F.2d 355	Yes	No
509 F.2d 293	No	Yes
514 F.2d 852	Yes	Yes
531 F.2d 364	Yes	Yes
532 F.2d 902	Yes	Yes
543 F.2d 395	No	Yes
555 F.2d 1046	No	Yes
559 F.2d 1251	No	Yes
562 F.2d 827	Yes	No
566 F.2d 696	Yes	No
578 F.2d 361	No	Yes
578 F.2d 880	Yes	No
580 F.2d 1331	Yes	Yes
584 F.2d 408	Yes	Yes
595 F.2d 897	Yes	Yes
598 F.2d 152	No	Yes
601 F.2d 33	Yes	No
601 F.2d 125	Yes	No
613 F.2d 1025	Yes	Yes
616 F.2d 65	Yes	No
619 F.2d 563	Yes	No
628 F.2d 36	No	Yes
628 F.2d 1283	Yes	No
631 F.2d 944	Yes	Yes
635 F.2d 891	Yes	No
657 F.2d 119	No	Yes
666 F.2d 454	No	Yes
675 F.2d 106	No	Yes
677 F.2d 22	Yes	Yes
683 F.2d 858	Yes	No
691 F.2d 953	Yes	Yes
692 F.2d 169	Yes	No
714 F.2d 588	Yes	Yes
727 F.2d 55	Yes	No
745 F.2d 76	No	Yes
745 F.2d 656	No	Yes
759 F.2d 922	No	Yes
760 F.2d 1297	No	Yes
789 F.2d 26	Yes	Yes
790 F.2d 1113	Yes	No
805 F.2d 1050	No	Yes
811 F.2d 149	Yes	Yes
819 F.2d 306	No	Yes
822 F.2d 1203	No	Yes
824 F.2d 332	Yes	No
844 F.2d 867	No	Yes
845 F.2d 345	No	Yes
855 F.2d 108	No	Yes
873 F.2d 884	No	Yes
888 F.2d 132	No	Yes
904 F.2d 172	No	Yes
909 F.2d 186	No	Yes
912 F.2d 1496	No	Yes
921 F.2d 313	No	Yes
938 F.2d 294	No	Yes
952 F.2d 500	Yes	No
953 F.2d 417	Yes	No
955 F.2d 254	Yes	No
955 F.2d 852	Yes	No
969 F.2d 1169	No	Yes
996 F.2d 122	No	Yes
997 F.2d 437	Yes	No
998 F.2d 1051	No	Yes
30 F.3d 169	No	Yes
41 F.3d 1300	Yes	No
41 F.3d 1532	Yes	No
47 F.3d 299	Yes	No
62 F.3d 1484	Yes	No
70 F.3d 1291	No	Yes
71 F.3d 574	No	Yes
84 F.3d 637	Yes	Yes
101 F.3d 772	No	Yes
104 F.3d 573	Yes	No
107 F.3d 882	Yes	No
115 F.3d 248	No	Yes
134 F.3d 125	No	Yes
140 F.3d 1085	No	Yes
156 F.3d 1010	Yes	No
168 F.3d 515	Yes	Yes
171 F.3d 478	No	Yes
193 F.3d 27	Yes	Yes
198 F.3d 139	No	Yes
198 F.3d 899	No	Yes
209 F.3d 760	No	Yes
222 F.3d 1030	Yes	No
234 F.3d 772	No	Yes
237 F.3d 683	No	Yes
250 F.3d 105	No	Yes
281 F.3d 235	No	Yes
282 F.3d 849	Yes	No
301 F.3d 167	Yes	Yes

Appendix C.

Summary Statistics.

Variable	Mean	10th percentile	25th percentile	Median	75th percentile	90th percentile	SD
Agency win (dummy)	0.53	0.00	0.00	1.00	1.00	1.00	0.50
Panel alignment with agency	0.01	−0.42	−0.20	0.00	0.26	0.42	0.31
Index of factual language	15.36	3.10	5.50	11.26	20.54	32.78	14.39
Index of legal language	23.01	5.22	9.71	17.58	31.85	47.45	18.38
Agency is appellant (dummy)	0.22	0.00	0.00	0.00	0.00	1.00	0.41
DC circuit (dummy)	0.26	0.00	0.00	0.00	1.00	1.00	0.44
Agency alignment with supreme court	0.00	−0.18	−0.09	0.00	0.09	0.20	0.14
Agency alignment with circuit	0.01	−0.37	−0.15	0.00	0.21	0.37	0.27

Declaration of Conflicting Interests

The author declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

Joseph L. Smith is associate professor of Political Science at The University of Alabama. He received his PhD from the University of Texas at Austin in 1999.

References

Aman

Mayton

(2001). Administrative law. (2nd ed.). St. Paul, MN: West Group.

Bartels

(2009). The constraining capacity of legal doctrine in the U.S. Supreme Court. American Political Science Review, 103, 474-495.

Baum

(1997). The puzzle of judicial behavior. Ann Arbor: University of Michigan Press.

Bowie

J. B.

Songer

D. R.

(2009). Assessing the applicability of strategic theory to explain decision making on the courts of appeals. Political Research Quarterly, 62, 393-407.

Brambor

Clark

Golder

(2006). Understanding interaction models: Improving empirical analyses. Political Analysis, 14, 63-82.

Breyer

Stewart

Sunstein

Spitzer

(2002). Administrative law and regulatory policy. New York, NY: Aspen Law and Business.

Brotherhood of Railway Carmen v. ICC, 880 F.2d 562 (D.C. Cir. 1988)

Coe

(2007). The language of freedom in the American presidency, 1933-2006. Presidential Studies Quarterly, 37, 375-398.

Corley

(2008). The supreme court and opinion content: The influence of parties’ briefs. Political Research Quarterly, 61, 468-478.

10.

Corley

Collins

Calvin

(2011). Lower court influence on U.S. supreme court opinion content. Journal of Politics, 73, 31-44.

11.

Corley

Howard

Nixon

(2005). The supreme court and opinion content: The use of the federalist papers. Political Research Quarterly, 58, 329-340.

12.

Cross

(2007). Decision making in the U.S. courts of appeals. Palo Alto, CA: Stanford University Press.

13.

Epstein

Martin

Segal

Westerland

(2007). The judicial common space. Journal of Law, Economics, and Organization, 23, 303325.

14.

Evans

McIntosh

Lin

Cates

(2007). Recounting the courts? Applying automated content analysis to enhance empirical legal research. Journal of Empirical Legal Studies, 4, 1007-1039.

15.

Epstein

Knight

(1998). The choices justices make. Washington, DC: Congressional Quarterly Press.

16.

Frank

(1951). Courts on trial. Princeton, NJ: Princeton University Press.

17.

Gibbons

(2003). Forensic linguistics: An introduction to language in the justice system. Malden, MA: Wiley-Blackwell Publishing.

18.

Gennaioli

Shleifer

(2008). Judicial fact discretion. Journal of Legal Studies, 37, 1-35.

19.

Giles

Hettinger

Peppers

(2001). Picking federal judges: A note on policy and partisan selection agendas. Political Research Quarterly, 54, 623641.

20.

Goldman

(1966). Voting behavior on the U.S. courts of appeals, 1961-1964. American Political Science Review, 60, 374-383.

21.

Goldman

(1975). Voting behavior on the U.S. courts of appeals revisited. American Political Science Review, 69, 491-506.

22.

Hall

Wright

(2008). Systematic content analysis of judicial opinions. California Law Review, 96, 63-122.

23.

Haire

Songer

Lindquist

(2003). Appellate court supervision in the federal judiciary: A hierarchical perspective. Law & Society Review, 37, 143-168

24.

Hume

(2009). Courting multiple audiences: The strategic selection of legal groundings by judges on the U.S. courts of appeals. Justice System Journal, 30, 14-33.

25.

Humphries

Songer

(1999). Law and politics in judicial oversight of federal administrative agencies. Journal of Politics, 61, 207-220

26.

Kastellec

J. P.

(2011). Panel composition and voting on the U.S. Courts of Appeals over time. Political Research Quarterly, 64, 377-391.

27.

Klein

Hume

(2003). Fear of reversal as an explanation of lower court compliance. Law & Society Review, 37, 579-606.

28.

Langer

(2002). Judicial review in state supreme courts: A comparative study. Albany: SUNY Press.

29.

Laver

Benoit

Garry

(2003). Extracting policy positions from political texts using words as data. American Political Science Review, 97, 311-331.

30.

Lindquist

Cross

(2005). Empirically testing Dworkin’s chain novel theory: Studying the path of precedent. New York University Law Review, 80, 1156-1206.

31.

Maltzman

Spriggs

Wahlbeck

(2000). Crafting law on the supreme court: The collegial game. New York, NY: Cambridge University Press.

32.

McGuire

Vanberg

(2005, September). Mapping the policies of the U.S. supreme court: Data, opinions, and constitutional law. Paper presented at the annual meeting of the American Political Science Association, Washington, DC.

33.

McGuire

Vanberg

Yanus

(2007, January). Targeting the median justice: A content analysis of legal arguments and judicial opinions. Paper presented at the annual meeting of the Southern Political Science Association, New Orleans.

34.

Monroe

Colaresi

Quinn

(2008). Fightin’ words: Lexical feature selection and evaluation for identifying the content of political conflict. Political Analysis, 16, 372-403

35.

Neely v. Shalala, 997 F.2d 437 (8th Cir. 1993)

36.

Owens

Wedeking

(2011). Justices and legal clarity: Analyzing the complexity of U.S. supreme court opinions. Law & Society Review, 45, 1027-1061.

37.

Perry

(1991). Deciding to decide: Agenda setting in the United States supreme court. Cambridge, UK: Harvard University Press.

38.

Primo

Jacobsmeier

Milyo

(2007). Estimating the impact of state policies and institutions with mixed-level data. State Politics and Policy Quarterly, 7, 446-459.

39.

Randazzo

Waterman

Fine

2006. Checking the federal courts: The impact of congressional statutes on judicial behavior. Journal of Politics, 68, 1006-1017.

40.

Richardson v. Perales, 402 U.S. 389 (1971).

41.

Segal

Spaeth

(1993). The supreme court and the attitudinal model. New York, NY: Cambridge University Press.

42.

Smith

Tiller

(2002). The strategy of judging: Evidence from administrative law. Journal of Legal Studies, 31, 61-82.

43.

Songer

Davis

(1990). The impact of party and region on voting decisions in the United States courts of appeals, 1955-1986. Western Political Quarterly, 43, 317-334.

44.

Songer

Ginn

Sarver

(2003). Do judges follow the law when there is no fear of reversal. Justice System Journal, 24(2), 137-161.

45.

Songer

Segal

Cameron

(1994). The hierarchy of justice: Testing a principal-agent model of supreme court-circuit court interactions. American Journal of Political Science, 38, 673-696.

46.

Songer

Johnson

(2002). The influence of presidential versus home state senatorial preferences on the policy output of judges on the United States district courts. Law & Society Review, 36, 657-676.

47.

State Corporation Commission v. Interstate Commerce Commission, 894 F.2d 1141 (10th Cir. 1989)

48.

Sunstein

Sawicki

Schkade

Ellma

(2006). Are judges political? An empirical analysis of the federal judiciary. Washington, DC: Brookings Institution Press.

49.

Zorn

Bowie

(2010). Ideological influences on decision making in the federal judicial hierarchy: An empirical assessment. Journal of Politics, 72, 1212-1221