The American Public’s Attitudes over How Judges Use Legal Principles to Make Decisions


 We investigate the American public's attitudes over an integral component of judicial behavior: the legal principles judges employ when making decisions. Our theoretical perspective argues that political preferences shape individuals’ attitudes over how judges apply legal principles, mirroring ideological divisions expressed by political elites and judges. Using an original battery of questions, we find high support across all Americans for the use of certain, well-established legal principles, but stark differences in how liberals and conservatives evaluate the use of more controversial principles. In a survey experiment, we find that agreement (disagreement) between an individual's attitudes over the use of legal principles and the reasoning contained in a Supreme Court opinion is associated with increased (decreased) support for the Court decision.


Appendix 1. The American Public's Attitudes over How Judges Use Legal Principles to Make Decisions Appendix 1.1 Real-World Elite Communication and Principles of Judging
In Table A-1 and Figure A-1, we use string searches to find direct references to a series of legal principles in press releases from U.S. Senators and network news transcripts about Supreme Court nominees, supplemented by manual coding to remove false positives.The search terms included variations in how to discuss a principle, such as "precedent," "stare decisis" and "settled law," or "judicial activism," "judicial restraint," and "legislate from the bench."Nevertheless, elite actors may also reference these concepts in forms our search terms could not capture, making this a likely undercount.This set of principles, while not exhaustive, attempts to capture a broad range of ideas surrounding judicial decision-making, including some of the most important principles that we asked about in our surveys (such as precedent and original intent) as well as direct references to key constitutional phrases and some other language used in popular discourse, such as "strict construction" and the "living constitution."We see in Table A-1 that this discussion by political elites is multi-faceted; Figure A-1, which focuses on news transcripts, highlights how such discussion appears across many nomination contexts.Figure A-2, which uses data from Collins and Ringhand 2013 and employs slightly different categories, shows that senators of both parties in the Judiciary Committee engage in rhetoric surrounding legal principles.

Appendix 1.2 Front Page Newspaper Discussion of Legal Principles
As we note in the main text, all 28 front-page New York Times stories on constitutional Supreme Court cases from 2010-2014 described the reasoning behind the decision given by the justices, often giving multiple types of rationales and noting the justifications used both by the majority and by dissenting justices.Front page cases noted in Table 2-13 of Epstein et al. (2015). 41For example, coverage of the Court's decision finding warrantless searches of cellphones unconstitutional in Riley v. California (2013) discussed the role of precedent ("the Supreme Court's precedents had supported the government"), historical notions of privacy ("One of the driving forces behind the American Revolution, Chief Justice Roberts wrote, was revulsion against 'general warrants' "), and societal consequences ("[Roberts was] keenly alert to the central role that cellphones play in contemporary life").See: https://perma.cc/43R7-JFUX.

Note:
The figure presents the proportion of network news and NPR segments about the last ten Supreme Court nominees that referenced that discussed principles of judging.Data come from 11,833 network news transcripts that mention one of the last ten Supreme Court nominees during the time period of their nominations cataloged by Nexis Uni.These transcripts come from ABC, CBS, CNN, Fox Business, Fox News, MSNBC, NBC, NewsHour, and NPR.In total, 20.4% of network news transcripts about Supreme Court nominees include at least one reference to the above principles of judging.

Appendix 1.6 Sample Details
The Harvard/Harris Poll uses an opt-in internet panel weighted to a target sampling frame based upon age, gender, region, race/ethnicity, marital status, household size, income, employment, and education.Respondents were also weighted by their propensity to be online and thus have the opportunity to take the survey.A possible concern about our question stem is that respondents will interpret the "importance" of a principle of judging not as a prompt about what the respondent thinks it is normatively important for judges to do, but as a descriptive question about what judges themselves say is important.We encourage the development of multiple ways of asking about principles of judging.However, we think that respondents likely interpreted the prompt in the way we intended.
First, to test whether the wording of our prompt may have led respondents to evaluate these questions in a different way than we intended, we included in our MTurk survey (see Appendix 1.15) an alternative wording, given to half the respondents, where we added "for judges to use when making decisions" when giving the four importance categories; this language is more clearly purposive ("important for") and thus appears less liable to misinterpretation. 42is change in wording did not meaningfully affect responses.Difference-in-means tests reveal no statistically distinguishable differences in the mean level of importance given to any of the ten principles questions across question groups (p-values range from 0.39 to 0.93). 43Additionally, the average absolute difference in finding a principle important (the sum of responses for "very important" or "somewhat important") between the groups is 1.3 percent.This increases our confidence that our respondents are interpreting our prompt in the way we intended.
Second, the differences that we find by ideology are more consistent with the view that respondents interpreted this question as normative rather than descriptive.The patterns we see, especially among highly knowledgeable respondents, track more closely the public rhetoric from liberals and conservatives about what the courts should do rather than the rhetoric about what they are actually doing.Furthermore, even though our question wordings differ significantly from the "judicial values" questions, we see some notable points of connection where there some topical overlap.Our question 42.This longer wording more closely resembles previous survey stems on "judicial values" (Gibson and Caldeira 2009b; Greene, Persily, and Ansolabehere 2011) which state "How important would you say it is for a good Supreme Court judge to. . ." (Greene, Persily, and Ansolabehere 2011, 366).
43.To test this, we include respondents who participated in another experiment, not discussed in this paper.Thus, our sample size is larger (872), reducing the likelihood that these non-results are a function of a small sample size.about precedent elicited 78% importance (Table 2) while we see 79% importance for "Respect existing Supreme Court decisions by changing the law as little as possible" in Greene, Persily, and Ansolabehere 2011, 366.Our question about original intent elicited 88% importance, while we see 92% importance for the somewhat related prompt of "Uphold the values of those who wrote our Constitution two hundred years ago" (Greene, Persily, and Ansolabehere 2011, 366).Finally, while our question about public opinion elicted much less support than their prompt "Respect the will of the majority of people in the United States" (48% importance versus 74% importance), we think that our explicit invocation of "current public opinion" will call to mind different and more negative considerations than "the will of the majority of people."Nonetheless, both surveys find less support for considering public opinion than for precedent or original intent.While our question wordings are novel and intended to capture different concepts from previous work, these broad points of connection also lead us to believe that our respondents were treating our prompt as normative and not descriptive.
Finally, if some respondents were interpreting our prompt as a descriptive question about what judges actually do, we should expect smaller ideological differences than if they were interpreting the prompt as normative.Thus, measurement error of this sort should bias our coefficients downwards, which would suggest that the true ideological differences in the population may be greater than we report.Note: Survey respondents were asked "What is your gender?" and response options were Male and Female.Because respondents were asked about gender, we use the term gender here, though we recognize the response options refer to biological sex.Appendix 1.10 Factor Analysis

SM -9 Appendix 1.9 Descriptive Results by Demographic Groups
Columns 1 and 2 of Table A-8 present the results from a principal components analysis with two components using a varimax rotation.Our analysis suggests two primary factors explain a considerable proportion of the variation in our data.A scree plot suggests two primary factors, as does a parallel analysis (see Figure A-3).The eigenvalue for the first factor is 3.29 (explaining 0.25 of the overall variance), and the eigenvalue for the second is 1.46 (explaining 0.22 of the overall variance).Respondent mean level of support for our traditional principles correlates with the first factor of a principal component analysis with two components at 0.88 and respondent mean level of support for our non-traditional principles correlates with the second factor at 0.93.
As our theory about traditional and non-traditional principles does not preclude the possibility that the factors could be correlated, we also conduct an exploratory factor analysis using the minimum residual method and an oblimin rotation.Columns 3 and 4 of Table A-8 show that the results do not differ greatly from the PCA.The first factor from this method correlates with average responses to traditionalism at 0.87 and the second factor correlates with average responses to non-traditionalism at 0.96.

Note:
The table presents factor loadings from an principal components analysis with two components using a varimax rotation (columns 1 and 2) and factor loadings from an exploratory factor analysis using the minimum residual method and an oblimin rotation (columns 3 and 4).Note: The left plot is a scree plot of the eigenvalues from our principal components analysis; the "elbow" appears at the third factor, suggesting retaining the first two factors.The right plot presents a parallel analysis comparing the eigenvalues from our analysis to randomly generated eigenvalues; the results similarly suggest retaining the first two factors.

SM -19
than in the October 2017 survey.
The Court may have been particularly salient at the time our study was fielded, given the allegations of sexual assault against then-nominee Brett Kavanaugh that emerged on September 12, 2018. 45Nevertheless, a few notable features of our study and findings help us remain confident that these events do not limit the external validity of our findings.First, our principles questions and experiment are either asked in the abstract or in the context of Supreme Court decisions, rather than as applied to nominees or individual judges.Second, legal principles were not the focus of popular attention and media coverage of the Kavanaugh nomination in this time period.Finally, the similar patterns in evaluations of principles between this survey and the October 2017 survey suggest that the Kavanaugh scandal did not meaningfully affect Americans' views of legal principles.
receiving "Neither Important" principles reduces support for the reasoning used in the case by 23 percent (p<0.03).This question was asked after the support question we use in our primary analyses.Table A-18 shows the results with Legitimacy as the dependent variable.Legitmacy is an additive index of four questions; each question is on a five-point scale, and we rescale the additive index from 0-1.The respondents indicated their levels of agreement with the following statements: "If the U.S. Supreme Court started making a lot of decisions that most people disagree with, it might be better to do away with the Supreme Court altogether," "The U.S. Supreme Court gets too mixed up in politics," "Judges on the U.S. Supreme Court who consistently make decisions at odds with what a majority of the people want should be removed from their position as judge," and "The U.S.
Supreme Court has become too independent and should be seriously reined in."

Figure A- 1 .
Figure A-1.Television News Coverage of Legal Principles

Figure A- 3 .
Figure A-3.Assessments of How Many Components to Retain

Table A -
1. Senator Press Releases about Principles of Judging McCain (R-AZ): "I know of no more profoundly anti-democratic attitude than that expressed by those who want judges to discover and enforce the ever-changing boundaries of a so-called 'living Constitution."OriginalIntent3.8%Biden(D-DE):"Judge Alito... has supported the theories of strict construction and originalism.He stated: I think we should look to the text of the Constitution and we should look to the meaning that someone would have taken from the text of the Constitution at the time of its adoption."Precedent11.3%Feinstein(D-CA): "The court could also overturn Roe v. Wade, stripping women of the right to control their reproductive systems.[Barrett]hasarguedthat justices have the right to ignore that precedent if they believe the Constitution doesn't protect a woman's right to choose."StrictConstruction0.8%Burns (R-MT): "[Alito] is a strict constructionist and deeply reveres the Constitution, and the Supreme Court will be well served by his membership."Textualism5.3%Portman(R-OH): "[Barrett] has a commitment to interpret the text of the Constitution and the laws as they are written, rather than through the lens of her own policy or personal preferences.I appreciate that modest approach.itleavesthe legislating to the representatives elected by the people rather than the unelected judges."Note:Tablepresents information on Senator press releases about principles of judging.Data come from all 3,694 press releases issued by Senators about the last 10 Supreme Court nominees, taken from ProQuest, senatorial websites, and Grimmer 2013.Categories are not mutually exclusive.The second column reports the percent of all press releases that reference the principle.In total, 25.2% of Senator press releases about Supreme Court nominees include at least one reference to the above principles of judging.Appendix 1.

Appendix 1.7 The Temporal Context of Our Survey
about legal principles in the context of his nominees differs from other elites.It is likewise possible that the attitudes we measure toward legal principles might differ under a different president or political environment.We see these questions as important avenues for future research.
(October 2017).President Trump emphasized the legalistic traits of his judicial nominees in his public communications (see, e.g., his comments on Neil Gorsuch presented in the main text).While popular discussion of legal principles in the context of judicial nominees and decisions is not unique to the Trump era (see, e.g., Figure A-1 above), it is possible that the ways in which Trump talked

Table A -
3. Percentage of Respondents Rating a Principle as Important by Gender

Table A -
4. Percentage of Respondents Rating a Principle as Important by Race

Table A -
5. Percentage of Respondents Rating a Principle as Important by Partisanship

Table A -
6. Percentage of Respondents Rating a Principle as Important by Education

Table A -
7. Percentage of Respondents Rating a Principle as Important by Income

Table A -
9. Politics, Judicial Knowledge and Support for Principles of Judging The tables present linear regression coefficients predicting support for each principle of judging as a function of knowledge of the Supreme Court and respondent ideology.The outcome variables capture evaluations of each principle's importance.The variables take on one of four values, 0, .33,.66 or 1, where higher values indicate greater importance.The Supreme Court knowledge variable ranges from zero to seven and is measured using a battery of seven questions; higher values indicate greater knowledge of the Court.* * * Note:The tables present linear regression coefficients predicting mean support for our categories of the principle of judging as a function of knowledge of the Supreme Court and respondent ideology.The outcome variables capture average evaluations of the importance of the principles in each category.Support for individual principles take on one of four values, 0, .33,.66 or 1, where higher values indicate greater importance.The Supreme Court knowledge variable ranges from zero to seven and is measured using a battery of seven questions; higher values indicate greater knowledge of the Court.* * * p < .01;* * p < .05;* p < .1.Note: p < .05;* p < .1.

Table A -
11.Politics, Judicial Knowledge and Support for Principles of Judging: Ordered Logistic Regressions The table presents ordered logistic regression coefficients predicting support for each principle of judging as a function of knowledge of the Supreme Court and respondent ideology.The outcome variables capture evaluations of each principle's importance.The outcome variables take on one of four ordered values, "not important," "a little important," "somewhat important," and "very important."The Supreme Court knowledge variable ranges from zero to seven and is measured using a battery of seven questions; higher values indicate greater knowledge of the Court.* * *

Table A -
12. Percentage of Respondents Rating a Principle as Important: MTurk Survey Table presents the percentage of respondents that evaluated each principle as very important, somewhat important, not very important, or not important at all for the September 2018 MTurk survey (N = 872).Values are unweighted.

Table A -
15. Support for Court Ruling as a Function of Judicial Principles The table presents the results of linear and logistic regressions predicting support for a case outcome as a function of the legal principles used to justify a decision and covariates; robust standard errors reported.Respondent-level covariates include race, gender, ideology, and the number of principles supported ex ante.The estimated directionality of the treatment effects, and in most cases the substantive significance, are consistent with the group means presented in the main text.

Table A -
16. Support for Court Ruling as a Function of Judicial Principles: Pooling Neither and Mixed The table presents the results of linear and logistic regressions predicting support for a case outcome as a function of the legal principles used to justify a decision and covariates; robust standard errors reported.Respondent-level covariates include race, gender, ideology, and the number of principles supported ex ante.The estimated directionality of the treatment effects, and in most cases the substantive significance, are consistent with the group means presented in the main text.

Table A -
17. Support for Reasoning as a Function of Judicial Principles * * * p < .01;**p < .05;*p< .1Note:Thetable presents the results of linear regressions predicting support for reasoning as a function of the legal principles used to justify a decision and covariates; robust standard errors reported.Respondent-level covariates include race, gender, ideology, and the number of principles supported ex ante.* * * p < .01;* * p < .05;* p < .1.

Table A -
18. Legitimacy as a Function of Judicial Principles The table presents the results of linear regressions predicting legitimacy as a function of the legal principles used to justify a decision and covariates; robust standard errors reported.Respondent-level covariates include race, gender, ideology, and the number of principles supported ex ante.