Scholarly article on topic 'Supreme Court Justices’ Economic Behaviour: A Multilevel Model Analysis'

Supreme Court Justices’ Economic Behaviour: A Multilevel Model Analysis Academic research paper on "Political Science"

0
0
Share paper
Academic journal
Scandinavian Political Studies
OECD Field of science
Keywords
{""}

Academic research paper on topic "Supreme Court Justices’ Economic Behaviour: A Multilevel Model Analysis"

ISSN 0080-6757 Doi: 10.1111/1467-9477.12060 © 2015 The Authors. Scandinavian Political Studies published by John Wiley & Sons Ltd

on behalf of Nordic Political Science Association.

Supreme Court Justices' Economic Behaviour: A Multilevel Model Analysis

Jon Kare Skiple,* Gunnar Grendstad,f William R. Shafferff and Eric N. Waltenburgfff

Supreme Court justices are overlooked, but important, national policy-making players who render final and consequential decisions in cases on economic conflicts. The research question asks what forces explain the decisional behaviour of Supreme Court justices in economic rights cases between a private and a public party. Theoretically, the decisional behaviour of an individual justice is a function of his or her notion as to what makes 'good' law, pursued in a cultural-collegial setting that is oriented by majoritarian requirements, while constrained by the legal nature of the case being considered. Empirically, all economic decisions made by Norwegian Supreme Court justices in five-justice panels from 1963 to 2012 are analyzed. Our multilevel model demonstrates that individual, collegial and case-level forces all contribute to explain the justices' votes. These results suggest that case-related dynamics, such as who the plaintiff is or the amount of disagreement between justices, matter, but also that ideology - via appointment mechanisms - matters when a nation's high court justices decide economic cases. Understanding the foundational assumptions and the institutional procedures is vital when transporting judicial behaviour models across polities.

Introduction

As Harold Lasswell (1936) put it nearly eight decades ago, politics is who gets what, when and how. Now, addressing these fundamental political questions is often reduced to left-right or public-private cleavages. This is certainly the case in legislative and electoral behaviour, but are these tensions and cleavages present in judicial behaviour in a nation's courts as well? The answer would seem to be but a matter of course. After all, a nation's courts make definitive rulings that result in winners and losers in

* Jon Kare Skiple, Department of Comparative Politics, University of Bergen, Christies gt.15, PB 7802, 5020 Bergen, Norway. E-mail: jon.skiple@uib.no

f Gunnar Grendstad, Department of Comparative Politics, University of Bergen, Norway ff William R. Shaffer, Department of Political Science, Purdue University, USA fff Eric N. Waltenburg, Department of Political Science, Purdue University, USA

This is an open access article under the terms of the Creative Commons Attribution License, which permits use, distribution and reproduction in any medium, provided the original work is properly cited.

the allocation of society's finite resources. Indeed, studies of judicial behaviour in the United States have found that left-right ideological preferences bear significantly upon judicial outputs.1 Political scientists in Europe, however, have been far less attentive to the possible effects of extra-legal forces, including ideological preferences, on judicial outputs.

In this article we attempt to partially remedy this omission. We analyze the decisional behaviour of Norway's Supreme Court justices in cases involving economic controversies, which is an excellent opportunity to test the transportability of a model of judicial behaviour developed and most fully documented in the American common law context. First, Norway's Supreme Court is no stranger to the distributional impact of economic cases. Examining all of the Court's decisions since 1963, we find that every fifth civil case pitted a public party against a private party over an economic conflict. Second, Norway's legal culture mixes common law and civil law legal systems (Sunde 2012b; Von Eyben 1956), and the Norwegian Supreme Court possesses several attributes that yield discretion to the individual justice's decisional behaviour. The Court sits at the apex of Norway's legal system. It has discretionary appellate jurisdiction, allowing it to choose cases with greater policy consequences. Finally, its members are appointed and sit for terms of good behaviour, largely freeing them from direct, democratic accountability.

Consequently, Norwegian justices may have a more active role in the making of law and policy than do their pure civil law counterparts. Indeed, the tension between the justices narrowly acting to resolve individual cases or to use those cases as a means to achieve other, broader ends traditionally has animated the Court. Law professor and future Chief Justice Carsten Smith's 1975 call for greater judicial activism is a case in point (Smith 1975). And by the mid-1990s, the policy makers prevailed: almost unilaterally the Supreme Court pronounced its mission to be 'uniformity, clarity and development of the law' (Sunde 2012b). The Court necessarily makes policy when developing the law, then, and depending upon the nature and indeterminacy of the case, Norway's justices are in a position to decide in a manner that is consistent with their policy preferences (see Segal & Spaeth 2002, 92-7).

To examine what factors influence justices' decisional behaviour, we compiled an original data file on all of the Supreme Court's decisions involving an economic issue from 1963 to 2012. This dataset comprises 806 unanimous and non-unanimous decisions and 4,000 justice votes. It is, to the best of our knowledge, the largest and most complete time-series on judicial-economic decision making in Europe. We then applied multilevel modeling to identify the individual, collegial and legal forces that bear upon the economic decision making of the 97 justices who served on the Supreme Court during this period. Although we discuss our findings in the context of the Norwegian Supreme Court, our research effort contributes to the emerging and

important study of judicial politics as a subfield of European comparative politics (Dyevre 2010; Volcansek 1993; Hanretty 2013).

Legal Traditions, Institutional Context and Judicial Behaviour

The Supreme Court

The Norwegian Constitution of 1814 established the Supreme Court as a court of last resort, consisting of a Chief Justice and at least six other associate justices (Clause 88).Today, there are 20 justices on the Court - one Chief Justice and 19 (associate) justices. The Court has discretionary jurisdiction, which is administered by its 'Appeals Committee'. This rotating three-justice committee screens all appeals and acts as the Court's gatekeeper, granting an appeal if any one of its members accepts it (Sunde 2012b).2 More recently and aided by the preparatory work of the Court's clerks, the Appeals Committee is only to accept a case if its outcome is expected to have an impact beyond the individual case itself. Consequently, each docketed case must be instrumental to the development of the law.

When the Appeals Committee accepts an appeal, the case is immediately docketed for oral argument with one of the two parallel rotating five-justice merits panels.3 The Chief Justice delegates to the Court's Director General the responsibility of allocating the 20 justices among the three-justice Appeals Committee and the two parallel five-justice panels. The quasi-randomized procedure ensures that the justices are distributed on each merits panel so as to maximize the range of seniority among the justices in order to guarantee that no panel consists of only junior justices and that an experienced justice (or the Chief Justice, if he is present) will chair each panel.

Following oral arguments, the five justices withdraw to deliberate and decide the case. First, the presiding justice offers his or her view on the case and how it should be decided. Then, in descending order of seniority, the other justices offer their views on their preferred outcome of the case. After stating their views, the justices cast their votes. Finally, one justice other than the presiding justice is selected to write the decision. If one or two justices disagree on the outcome of the case or intend to offer concurrences, opinion authorships may be redistributed (Schei 2010).

Modelling Judicial Behaviour

We analyze the justices and the votes they cast within the institutional setting just described. Theoretically, we propose a goal or motivation-

oriented model of the decisional behaviour of an individual Norwegian Supreme Court justice (see Baum 1997; Epstein & Knight 2013) - that is, the individual justice's vote is a function of his or her preferences for making 'good' law, pursued in a collegial setting that is oriented by majoritarian requirements, while constrained by the nature of the appeal being considered. Thus, individual attitudes and attributes, the presence and preferences of other justices, and the legal elements of the appeal under review shape and constrain how a justice decides. In other words, socialization plays a role in judging on the Norwegian Supreme Court. Institutionalized norms and rules and enduring social interactions influence the individual justice's decisional behaviour (see Beyers (2010,909) for a definition of 'socialization').

This socialization results in factors affecting judicial behaviour being interrelated in complex ways. We agree with Gibson's (1983, 8) argument that a theory of judicial behaviour 'can best be developed through models that incorporate influences stemming from various levels . . . but that are ultimately focused on the individual' since a justice casts his or her vote in order to settle a dispute. Accordingly, we muster hypotheses pertaining to economic decision making to empirically test the degree to which the justices' decisional behaviour is influenced by their ideology, background, collegial setting and the facts of the case. In the absence of well-established European-based theories of judicial behaviour and in following scientific pursuits of both disconfirmation and exploration, the hypotheses we offer are necessarily both deductively and inductively driven. We test our hypotheses in three equally important categories: attitudinal and attributive effects, panel effects and case effects (see Table 1).4

Justices as Individual Decision Makers: Attitudes and Attributes

To begin, we identify a set of justice-level forces. Similar to theories on governmental decision making in other non-majoritarian government institutions that follow a meritocratic recruitment procedure, we expect that demographic characteristics of judges are important for understanding their decisional behaviour (see, e.g., L^greid & Olsen 1978; Christensen et al. 2001; Trondal 2007). These forces are primary and secondary socialization effects, and as such we expect, to varying degrees, they will bear upon the decisional behaviour we are attempting to model.

According to the attitudinal model (Segal & Spaeth 2002), justices arrive at decisions that are consistent with their ideology and attitudinal preferences that inform their understanding both of the facts of the case and legal sources. Hence, different justices presented with the same legal facts arrive at different conclusions because different preferences informed their decisions. This explanation of judicial behaviour has found traction in settings beyond the United States. Hönnige (2009) in his study of the German and

Table 1. Variables, Operationalization, Hypotheses and Descriptive Statistics: The Norwegian Supreme Court, 1963-2012

Hypothesized

Variable Operationalization relationship Mean Sx Minimum Maximum

Dependent variable

Justice's vote 0 = Voted for private party 0.59 0.49 0 1

1 = Voted for public party

Justice level (v = 13)

SocDemGov appoint 1 = by soc.govt; + 0.68 0.47 0 1

0 = by nonsoc.gov't

Private practice 1 = Private practice as - 0.37 0.48 0 1

lawyer

Law professor 1 = Former law professor -/+ 0.07 0.26 0 1

Earlier judge 1 = Former lower court -/+ 0.48 0.50 0 1

Government advocate 1 = Former service + 0.22 0.41 0 1

Legislation department 1 = Former service + 0.40 0.49 0 1

Public prosecutor 1 = Former service + 0.09 0.28 0 1

Woman 1 = Woman; 0 = Man -/+ 0.17 0.37 0 1

Oslo born 1 = Born in Oslo (or + 0.41 0.49 0 1

Kristiania)

Age at vote (C) Justice's age at voting, -/+ 0.00 6.38 -20 10

centred

Seniority Justice's seniority in years -/+ 8.78 6.64 0 29

Chief 1 = Chief -/+ 0.04 0.20 0 1

Interim justice 1 = Temporary appointee + 0.03 0.18 0 1

Panel level (v = 8) (Panel majority consisting

of . . .)

SocDemGov majority 1 = Soc.gov't appointees + 0.81 0.39 0 1

majority

Priv. practice majority 1 = Justices w/private - 0.25 0.43 0 1

practice majority

Law professor majority 1 = Law professors (max = -/+ 0.01 0.10 0 1

3) majority

Gov't advocate majority 1 = Gov't advocate justices + 0.14 0.34 0 1

majority

Leg. dept majority 1 = Gov't leg. dept justices + 0.30 0.46 0 1

majority

Women majority 1 = Female justices - 0.06 0.23 0 1

majority

Oslo majority 1 = Oslo born justices + 0.32 0.47 0 1

majority

Chief in panel 1 = Chief is presiding -/+ 0.21 0.41 0 1

justice in panel

Case level (v = 7)

Public party plaintiff 1 = Public party is the - 0.38 0.49 0 1

plaintiff

Third parties N legal third parties -/+ 0.08 0.56 0 13

(intervenients)

Words N words in decisionb -/+ 3209.15 1709.57 391 11,864

Voices N additional opinions or -/+ 1.31 0.55 0 4

concurrances

Nonunanimous decision 1 = The decision is - 0.28 0.45 0 1

non-unanimous

Economic relevance high 1 = High economic -/+ 0.39 0.49 0 1

relevance

Economic relevance 1 = Medium economic -/+ 0.22 0.42 0 1

medium relevance

Notes: The data file includes 806 decisions and 4,000 justice-vote observations. aPanel majority refers to three or more justices of the relevant category on the panel. Age and seniority were not averaged at the panel level. Justices from the public prosecutor, interim justices and earlier justices were never in panel majority in the cases we study here. bNumber of words in decisions omits any concurrances or dissents.

French supreme courts, as well as Magalhaes (2003) in his study of the constitutional courts in Spain and Portugal, conclude that the justices' ideology predicts their votes. Hanretty (2012) shows that judicial dissent on the Spanish and Portuguese Constitutional Tribunals can be explained by their ideal points along a left-right dimension. Voeten (2008) finds that some judicial behaviour on the European Court of Human Rights (ECHR) is attributable not to legal culture, but to the preferences of their appointing governments. And Grendstad et al. (2011) detect similar attitu-dinal effects in non-unanimous economic decisions on the Norwegian Supreme Court.

In the absence of a direct measure of a justice's ideology, we use the nature of the appointing government as a proxy (Dyevre 2010). This measure assumes that there is an expected sharing of preferences between the government of appointment and the appointee in that the government in the winnowing process may choose or attract nominees who somewhat share the policy preferences of the appointing power. Specifically, we hypothesize that justices appointed by social-democratic governments and non-socialist governments are more likely to vote for the public and private party, respectively.

We argue that critics of the sharing-of-preferences hypothesis fail to demonstrate that this winnowing process cannot take place. Smith (1993) pointed to a flaw in the Norwegian system in which political appointments were possible.5 Moreover, three incidents of sharing-of-preferences underscore the validity of the appointment hypothesis. Former Labour Party Minister of Justice Jens Haugland writes candidly in his memoirs on politicians' preferences for candidates to the Supreme Court (Haugland 1986). Former conservative Prime Minister Kare Willoch takes pleasure in the non-socialist appointments of justices who were instrumental in securing the narrow conservative victory in the 1976 landmark Kl0fta decision (Willoch 2002). And in 1991, the social-democratic government of Gro Harlem Brundtland brushed aside the Supreme Court's own candidate for Chief Justice and instead awarded the seat to law professor Carsten Smith (Grendstad et al. 2015). Therefore, we assume a sharing-of-preferences mechanism similar to Adolph's (2013,263) demonstration of appointments to central banks: '[L]eft-leaning governments tend to pick central bankers whose career backgrounds foster dovish monetary policy preferences, whereas right-wing governments choose central bankers whose careers identify them as likely inflation hawks.'

The justices' occupational pursuits have secondary socialization effects. They engender experiences, relationships and memberships in social networks that produce or reinforce worldviews and attitudes. We test six variables that tap previous occupation. Following Tate (1981), we expect that justices who are recruited from private practice are more likely to relate to

the elite, conservative, socioeconomic interests that they had previously represented. Consequently, they would be more likely to find for the private party. Kj0nstad (1999) suggested a 'government-friendly hypothesis' in which justices with a career background in government offices are more likely to side with the public party, at least when it comes to civil cases. Therefore, we include measures of prior experience in the offices of the Director General of Public Prosecutions, the Legislation Department of the Government's Ministry of Justice or the Government Advocate. Finally, we also test the occupational effect of lower court judges and law professors. Both occupations have served the public interest of law, order and education, and both involve previous public sector employment. Whereas lower court judges have been recruited to the Supreme Court on a regular basis throughout the time period our data cover, it was not until 1991 that former law professors joined the Court. Although we expect an effect, we are unable to offer directional hypotheses for these two variables.

The internal organization of the Court imposes unequal roles and leverage among the justices. The Chief Justice is always the presiding officer of any panel on which he sits and always speaks first during deliberations. Five justices have served as Chief Justice during the period we study, and we are unable to venture one unidirectional hypothesis for them. Interim justices, on the other hand, occupy the bottom rung on the Court. They know that if a vacancy occurs, the power of permanent appointment lies in the hands of the government. Consequently, they are in a sense auditioning, and we venture that they more often than not will vote in deference to the public party. Finally, there may be some acclimation effects on a justice's decisional behaviour. Accordingly, we examine the effect of a justice's seniority on decisional behaviour (on seniority's effect with respect to socialization, see Trondal (2007, 1118)). We are unable to hazard a directional hypothesis regarding its effect on economic voting, however.

We also explore the effect of three primary socialization effect variables -age, place of birth and gender. We add the justices' age as a necessary control. The Norwegian centre-periphery cleavage (Rokkan 1967) would suggest that Oslo born justices would perceive the public-private distinction differently from those who were born and raised in the periphery (see McGuire 1993). A limited study on economic voting found Oslo born justices to decide for the public party (Grendstad et al. 2011). Although a Canadian study indicates that female justices are more left-leaning in private economic cases (Johnson et al. 2011), recent empirical evidence casts sufficient doubt on the causal connections between a judge's sex and his or her decisional behaviour (see Boyd et al. 2010) for us to venture a directional hypothesis.

Justices as Socially Embedded Decision Makers: Panel Effects

The second category of hypotheses captures the justices as decision makers constrained by legal culture and their shared social environment. Evidently, tension exists between the court as 'a collegial body' and its constituent justices - with the latter never being able to escape the Court's 'collegial coercion' (Michalsen 2006, 90). For the United States Supreme Court, Maltzman et al. (2001) demonstrate how the positions of the other justices constrain the individual justices' decisions. Epstein and Knight (1998) show how justices realize that their ability to prevail depends on the preferences of the other justices. Dyevre (2010,302-4) surmises that collegial deliberations may be stronger in European courts, but that their effects may be harder to measure.

We interpret panel effects as occurring when a majority or all of a panel's justices with similar socialization experiences vote in the same direction. A subcategory includes when a majority sways the minority of justices to vote in the same direction. In other words, we control for the possibility that certain types of panel majorities affect a justice's decisional behaviour. We can observe these effects indirectly by studying the rotation of justices on panels. Formally and all else being equal, we test for panel effects in that justice x will vote differently in panel y than in panel z (see Farhang & Wawro 2004; Sunstein et al. 2006; Meinke & Scott 2007; Kastellec 2011).

Given the nature of our individual-level variables, we are only able to hypothesize panel effects through majority measures for eight of these 13 variables.6 We expect that justices are more likely to find for the public party in economic cases when in panels with majorities of social democratic appointees, former government employees and justices born in Oslo. Further, we expect justices who vote in panels with majorities of justices with former private practice to be more likely to vote in favour of the private party. Based on an initial study by Skiple (2015), we anticipate that justices serving on panels with a majority of female justices are more likely to find for the private party. We are unable to hazard hypotheses concerning the panel effects of law professors or of the Chief Justice chairing the panel.

Justices as Legally Constrained Decision Makers: Case Effects

The third category of hypotheses speaks to the jurisprudence position that justices are not free to decide a case in any way they prefer. Simply put, the legal sources and 'the nature of the individual case' before them (as legal scholars are prone to say) constrain the justices. So in addition to limiting our study to economic cases, we also include properties of the case in order to compare the explanatory power of 'legal and non-legal forces' (as political scientists are prone to say). We deal with the some-

what elusive nature of the individual case in two ways: we control for the unobserved properties of the individual case through our two-level model (see the method section below); and we include seven measures of case properties.

It is perhaps to state the obvious that an appellant hopes the justices will overturn the lower court's decision, and there is some evidence that courts with discretionary jurisdiction accept cases to reverse them. Thus, all things being equal, the appellant might be advantaged.7 A somewhat different logic may be at work when the public party is the appellant. In these instances, the public party must consider the chances of its appeal passing the Appeals Committee. It must also weigh its resourcefulness as a litigant and the principle question of the case against the odds of it losing (Fagernas 2007). If the public party has lost in a lower court, it may still decide to appeal even when that calculus seems to recommend against doing so. The public party's decision to appeal may stem from its desire to reach closure on the current legal question: certainty of the law trumps its interest in winning the instant case.8 So, if the public party is the appellant, we hypothesize that the justices will still be more likely to find for the private party since overturning a lower court's decision is less likely than upholding it.

A primary characteristic of a case is its complexity. Complex cases are legally intriguing, but harder to resolve. We identify complex cases through the number of third parties supporting a litigant (i.e., legal intervenient), the number of justices who voice their opinion and the number of words in the majority opinion. Closely related to complexity is dissent among the justices. Indeed, dissent might even be seen as something of a rough proxy for complexity (Bentsen 2014). After all, complex cases present justices with greater opportunity to voice their principled position or pursue their goals. Clear legal provisions and elements, on the other hand, guide simple or 'easy' cases. Regardless of the individual justice's preferences, there should be but one legal solution to the easy case, and as a result, there will be no disagreement. 'The very hard [complex] case gives judges free rein to justify whatever decision best accords with their policy views' (Baum 1997,66). To the degree that support for the government or public party is a 'standing decision', then, dissents indicate the presence of complexity and the opportunity for the individual justices to vote their policy preference, to move away from the standing decision - simply put, to vote for the private party.9

The economic issue of the case separating the public and private party may be more or less salient to the case. If economic saliency matters, justices may adjust their votes depending on how much is at stake in the case. But if economic cases are homogeneous and offer similar cues, then judicial behaviour would be uniform across the cases. Therefore, all cases were coded for low, medium or high economic relevance; we use low economic relevance as a reference category in the analysis.10

Data, Method and Procedures

Data and Variables

Two Court decisions illustrate the type of economic cases analyzed here. In a 1985 tax case, the Court decided unanimously in favour of the state, represented by the Ministry of Finance, that a city-based business man who resided on his farm only during the summer could not deduct the deficit of his farm business from his other substantial business income since the farm was organized as a residential retreat and not a business (Rt-1985-319). In 2004, a petroleum tax case concerned for which fiscal years Shell, an international oil company operating on the Norwegian continental shelf, could deduct costs after closing depleted oil fields. Three of the justices voted in favour of Shell, while two justices voted in favour of the state, represented by the Ministry of Finance (Rt-2004-1921).

We searched the judicial Lovdata.no database for all Supreme Court decisions in five-justice panels since 1963 that addressed an economic issue and that pitted a public party against a private party. Three issue areas dominate: tax cases, expropriation cases and tort cases. We retained for analysis the 806 cases that identified a clear winner in either a unanimous (N = 579; 71.8 percent) or a non-unanimous (N = 227; 28.2 percent) decision. The 806 decisions represent 21 percent of all the civil cases the Court decided between 1963 and 2012.

A total of 97 individual justices, including 18 interim justices, are represented in our data. The 79 permanent justices on average cast 49.3 votes (Sx = 26.1; Range 1-107), while the interim justices averaged 4.5 votes (Sx = 3.9; Range 1-15). Because of missing values among the earliest justices, our final data set consists of 4,000 justice-vote observations.

Our dichotomous dependent variable measures whether a justice votes in favour of the public (= 1) or the private party (= 0) in the case. A public party includes the state, a county, a municipality or a publicly owned business. A private party includes an individual, a group of individuals or one or more private companies of any size. Descriptive statistics for the variables are shown in Table 1.11

Judicial Decision Making as a Two-level Regression Model

Regression models build upon the assumption of independent observations. However, our judicial behaviour data contain two potential sources of dependency between votes as the observational unit - namely the case and the justice. First, social life, as interactions between individuals and their contexts, applies to judicial life on the Norwegian Supreme Court too. Since the justices vote in rotating five-justice panels on a specific case with specific case facts, it is more likely that the justices' votes will correlate more within

cases than between cases. Second, since justices may vote many times and at different points in time, one could consider our data unbalanced repeated measure data. Consequently, we could also expect within-justice dependency because the same justice may vote similarly over time either in favour of the public party or in favour of the private party because he or she is predisposed to vote in certain ways.

This decisional context implies that our data structure is hierarchical because votes are nested within two higher groups of justices and cases, and cross-classified because there is a non-hierarchical relationship between cases and justices since the same justice votes in different cases. If we ignore the hierarchical dependency part of this data structure, the standard errors will be inaccurate (Hox 2010,5). If we ignore the cross-classified part of this data structure, between-group variance components may be biased (Luo & Kwok 2009).

The structure of our data, then, suggests a cross-classified two-level model where the justice votes at level 1 are nested within the case and justice at a parallel level 2. This cross-classified multilevel model, however, is complex and computationally demanding. Therefore, we follow Hox's (2010) suggestion and estimate preliminary models in order to empirically assess if within-case-dependency and within-justice-dependency exist, and we let the results from these tests guide our model selection. We estimate the dependencies in the data through the intra-class correlation, which measures the implied correlation (i.e., similarity or homogeneity) of the observed responses within a given group. In our data the intra-class correlation coefficient of votes within justices is 0.018 and of votes within cases the equivalent estimate is 0.920. In other words, the within-case-dependency is very strong, while the within-justice-dependency is relatively weak. Although we should not neglect the within-case-dependency, the data suggest that we are permitted to neglect the within-justice dependency and, in turn, also the cross-classified data structure.

Therefore, we employ a two-level model where the justices' votes are nested within cases. The two-level model consists of: a dependent variable that captures the individual justices' votes; explanatory variables fixed at both the justice level and the panel-cum-case level; and a random intercept for each case. We treat the panel as a characteristic of the case, thus placing panel variables as a set of fixed effects at the case level (see Fielding 2010).

This two-level model provides three advantages. First, the model accounts for the dependency between votes within the same case. This also implies that our model controls explanatory variables for the 'average case' - that is, the case with the average probability for the public vote. Thus our two-level model effectively rebuts critics who claim that judicial politics scholars neglect the individual case in their statistical analyses (see Sunde 2012a). Second, the model allows us to explore the dependency in the observations

through the intra-class correlation in the random part of the model. In addition to the implied correlation between observations (i.e., justice votes) within the same group (i.e., case), the intra-class correlation can be understood as the share of group-level variance compared to the total variance. Third, the model allows us to employ our theoretical framework in an efficient manner. Variables pertaining to different levels of analysis can be analyzed at the correct and appropriate level simultaneously. In so doing, we avoid deductive and statistical problems linked to aggregation and disag-gregation of data (Hox 2010,3). In other words, the theoretical implications of the model is that the individual level variables (e.g., woman) can explain the within-case variation in the data (i.e., non-unanimous decisions), while the panel level variables (e.g., a majority of women on the panel) and the case-level variables (e.g., public plaintiff) can explain the between-case variation in the data.

Since our dependent variable is dichotomous, assumptions of continuous scores, normality and homoscedastic errors are violated. Therefore, we employ a generalized linear model with a logit-link function and the binominal Bernoulli distribution. We report the regression coefficients in our models as log-odds. We transform the log-odds to predicted probabilities in order to assess the substantive effect. In generalized linear models the lowest level residual variance (i.e., within-case variance) does not enter the model as a separate term, but is estimated to be 3.29.12

Procedures

We begin by estimating an intercept-only model in order to test empirically whether the theoretical and statistical arguments for our two-level model hold. At three intermediate and sequential steps (see the Online Appendix) we enter and test all variables pertaining to the individual, panel and case levels. At each step, on the basis of their contribution or lack thereof, we determine whether a variable merits continued inclusion. We excluded insignificant variables according to the rule of parsimony and to facilitate model estimation. Given our population data of economic behaviour on the Supreme Court, we retain an independent variable in the model when its effect knocks on the p < 0.10 level of statistical significance, but we confirm hypotheses at the p < 0.05 level only. We now turn to our two-level model that explains judicial votes in all economic cases for the 1963-2012 period.

Results: Justices' Economic Voting Explained

A Two-level Model

The intercept-only model (see Table 2) provides two crucial statistics. First, the model offers strong evidence that the between-case variance is non-zero

Table 2. The Intercept-only Model - Votes Nested in Cases: The Norwegian Supreme Court, 1963-2012

B Standard error P > Izl

Intercept 1.56 0.27 ***

Intra-class correlation 0.92

Explained variance

AIC 3,055.81

BIC 3,068.40

N (Decisions) 806

N (Votes) 4,000

N (Justices per case) 5(4)

Notes: Two-level logistical regression model. Likelihood-Ratio test of rho = 0; Chibar2 = 2,357.91; Prob >= chibar2 < 0.001. *** p < 0.01.

(chibar2 = 2,357.91; p < 0.001), which indicates that a two-level model is superior to a single-level model. Second, the intra-class correlation demonstrates that the case structure of the data accounts for 92.1 percent of the variance in our dependent variable. This means that the variance in votes is due largely to differences between cases at the panel and case levels, rather than differences within cases at the individual level. This result is not surprising given the dominance of unanimous decisions (71.8 percent) in our data.

Forces Impacting the Justices' Economic Decision Making

Table 3 conveys the essence of the Supreme Court justices' economic decision making for the 1963-2012 period.13 Our results demonstrate a parsimonious model of judicial behaviour in the important field of economic decision making. First, there are two significant forces on justices as individual decision makers. Social-democratic government appointees are more likely to vote for the public party than are non-socialist appointees. In more concrete terms, social-democratic appointees are 7 percent more likely to find for the public litigant.14 Prior experience as a law professor seems to draw these justices away from voting for the public party. The reduced probability for this group is 22 percent. At least for this set of justices, their previous work experience systematically shapes their behaviour on the bench.

Second, there is one significant force on justices as socially embedded decision makers: when a panel consists of a majority of female justices, the probability of all the justices voting for the public party decreases by 43 percent. Although there is not an appreciable effect of gender at the justice level, the collegial effect of gender kicks in strongly at the panel level.

Table 3. Justices' Private versus Public Economic Voting on the Norwegian Supreme Court, 1963-2012

B Economic voting Standard error P > Izl

Justice level

Social-democratic gov't appointee 0 47*** 0.16 0.003

Law professor -1.22*** 0.28 0.000

Public prosecutor 0.43* 0.26 0.096

Seniority -0.20* 0.01 0.068

Panel level

Social-democratic gov't appointee majority 0.97 0.59 0.102

Female justices majority -2.07** 0.98 0.035

Case level

Public party is plaintiff -4.35*** 0.51 0.000

Non-unanimous decision -2 42*** 0.49 0.000

Constant 3.25*** 0.61 0.000

Intra-class correlation 0.912

Explained variance (AIC/BIC) 2938.45 3001.39

Notes: Two-level logistical regression model. Dependent variable is justices voting for the private (= 0) or public (= 1) party in economic cases. The coefficients show the effects of the forces on voting as log odds. *p < 0.10; **p < 0.05; ***p < 0.01. N (Votes) = 4,000, N (Cases) = 806.

Evidently, when in a majority on a panel, elite female justices are able to sway their colleagues to side with the private party. It also bears mentioning that there are no 'legal' forces at the panel level, such as effects of legal careers or status on the court.

Third, there are two significant case-level forces. First, when the justices decide an economic case where the public party is the plaintiff, the likelihood of a justice finding for the public party drops by 70 percent. This strong tendency of individual justices to rule against the public party when the public party is the source of the appeal suggests that an expectation of success does not figure prominently in the public party's litigation strategy. As we suspected, achieving clarity of regulations and administrative rules with respect to economic issues is of greater importance than a finding in favour of the public party litigant. Second, when there are dissenting justices present on a panel, there is a reduced probability of 43 percent that all the justices on the panel will vote in favour of the public party. Simply put, justices are more likely to dissent in favour of the public party. Our hypothesized effect of complex cases leading to dissents and a move away from the pro-public party standing decision appears to have some empirical support.

Disconfirmed Hypotheses

A total of 23 of the 28 independent variables that were hypothesized as bearing upon the justices' economic behaviour failed to attain significant

effects. Furthermore, 11 of the 13 individual-level variables failed to gain any traction. Evidently, a larger number of justice's attributes are irrelevant to the justices' votes. And this seems to be consistent with the meager traction primary socialization effects have on accounting for decisional behaviour in governmental institutions (Christensen et al. 2001). This finding, to a large extent, seems to mute the claims of the attributive theory of judicial behaviour. Similarly, seven of the eight panel-level variables had insignificant effects. Again, a large number of collegial features failed to show any signs of affecting the justices' voting. Finally, five of the seven case-level variables fell by the wayside. Case complexity - words, voices or parties - failed to account for judicial voting patterns; nor did variation in the economic saliency of the case move the justices.

Discussion and Conclusions

Foundational Assumptions

According to Baum (1997,13), justices, like people in general, 'have sets of goals that they would like to achieve'. In the case of justices, these goals comprise such things as the correct interpretation of the law, the creation of legal policy, embellishment of reputation, career advancement, a manageable workload and good collegial relations (see Baum 1997, Table 1.1). Baum further avers that in order to understand judicial behaviour, one must recognize that the achievement of these goals is a function of the institutional and legal situation in which a justice is acting. In other words, a skein of forces constrains and directs the individual justice's behaviour. Judge Posner describes this complex situation nicely:

[A judge's] response to a case is generated by legal doctrine, institutional constraints, policy preferences, strategic considerations, and the equities of the case, all mixed together and all mediated by temperament, experience, ambition, and other personal factors. (Posner 2008, cited in Epstein & Knight 2013,25).

Our application of this understanding of judicial behaviour to Norway's Supreme Court justices gave rise to a number of expectations concerning the forces that systematically affect how they act. First, we hypothesized that individual-level forces influence a justice's decisional behaviour. Our notion here is that ideology is one of several individual-level motivations that affect a justice's decisions. In addition, we suggested (and controlled for) the effects of a number of demographic attributes and socialization experiences that might systematically bear upon a justice's decisional behaviour. In each case, it was not the attribute or experience per se that was expected to affect behaviour, but the attitudes that they likely gave rise to that would influence judicial outputs.

Second, the institutional context of the Norwegian Supreme Court provides another set of forces to include in our model specification. The Supreme Court is a majoritarian institution, and the great bulk of its decisions are the product of five-justice panels. Consequently, we expected panel effects. Specifically, we controlled for the possibility that certain types of panel majorities, or legal culture, affected a justice's decisional behaviour. Here, the notion is that decisional majorities may exercise something akin to a gravitational pull on recalcitrant justices, bending their votes in the same direction as the majority. This might occur because the recalcitrant justice wants to maintain good collegial relations, limit in some manner his or her workload, or because the minority justice(s) are won over by ghosts of legal culture past or the unique and especially credible information upon which the majority is acting (Baum 1997; Boyd et al. 2010,392; Epstein & Knight 2013).

Finally, it is an individual actor's operational behaviour (i.e., pursuit of goals; see Baum 1997) in a court of law that we are attempting to explain. As such, our analytical model must account for forces that bear upon outcomes that largely are beyond the individual justice's control (see Epstein & Knight 2013,24). By their nature, courts are reactive institutions. Thus, the individual justice has very little say over the specific elements of the cases that act as the vehicles for the pursuit of his or her goals. The justices do not determine a priori the clarity of the law at issue in a case brought to them. They do not manage the nature or differential status of the litigants in the appeal. And they do not establish the complexity or salience of the legal questions brought to the bar. Each of these case attributes constrains a justice's behaviour. Accordingly, we included their measures in the specification of our multilevel model.

Main Findings

Elements from each category of our estimated model possess some statistical traction. Thus, our findings are consistent with a goal-oriented interpretation of judicial behaviour (see Baum 1997; Epstein & Knight 2013). Interestingly, inter-justice variation takes a back seat to panel and case-level factors in explaining a justice's decisional behaviour. This is not to say, however, a justice's policy preferences have no role. Indeed, attitudinal preferences may have a certain primacy in a justice's motivational ordering, inasmuch as they can be at the root of the achievement of many other goals. (Job satisfaction, for example, is much more likely if one gets one's way.) But our model shows that attitudes are not the only force affecting behavior - an outcome that echoes a point made by Epstein and Knight: An exclusive focus on ideology is not valid because 'ideological motivations are just one of several kinds of motivations that should be incorporated into a realistic

and comprehensive conception of judicial decision making' (Epstein & Knight 2013, 24).

Limitations of the Study

Now, cobbling together a 'more realistic and comprehensive' model of judicial behaviour is no mean feat, and the model we specified and estimated above has several notable limitations. First, as we noted above, the intra-class correlation statistic derived from the multilevel modeling procedure indicates that the lion's share of the variance in our dependent variable is a panel or case-level phenomenon. No doubt this is related to the preponderance of unanimous cases in our dataset (71.8 percent). Consequently, there is relatively little variance left to be explained by individual, justice-level forces. Until we are better able to measure and include forces that distinguish across cases and panels in our specification, our model's explanatory power will remain low.

Second, more valid and precise measures of the justices' various goals and motivations are needed. This is particularly true for goals other than those relating to policy preferences, although our use of the colour of the appointing government as a crude proxy for a justice's ideology could certainly be improved upon. Our model indicates that panel and case-level effects bear upon behaviour, but to a large degree our conclusions along these lines are the product of circumstantial evidence. To be sure, goals concerning workload and collegiality seemingly are consistent with a justice's decision to join the majority, but without direct soundings on the salience of these goals to the individual justice, we are not able to determine the conditions under which they take on greater or lesser weight. Similarly, case attributes matter with respect to a justice's motivation to 'follow the law', but our current model is largely devoid of such measures. Here, too, then, we would be well advised to derive instruments that tap such case attributes as 'precedents, constitutional provisions, statutes, administrative regulations, legislative history, ... social practices ..., social norms, and basic values' and their effect on judicial interpretations and outcomes (Epstein & Knight 2013,25).

Finally, 50 years of data on economics decisions almost guarantees the introduction of temporal variation in public attitudes toward the role of government in the economy. A model more sensitive to the effect of changes across time might enjoy greater explanatory power. Furthermore, such a model might offer clues as to whether the Court lags or leads in the development of the law in an issue domain of such societal consequence. Temporal variation may also be in effect for the mix and relative weights of the different motivations influencing the justices' decisions. In the case of the United States Supreme Court, for example, George and Epstein (1992,334) have shown that legal forces (i.e., certain case attributes) have 'the greatest

impact at the early stages of an issue's life; as it evolves, however, extralegal variables [such as preferences] dominate'.

Comparative Implications

The Norwegian legal system is comparable to, and our results may be tested in, the other Nordic countries, in that the former East-West legal distinction that might have militated against comparisons in this area now seems to have vanished. Additionally, our model of judicial behaviour may be transportable to other European legal systems. These systems have responded to the significant increase in the number of appeals by accepting fewer and more policy consequential cases rather than by adding justices. As a result, the justices are better able to treat the cases before them as instruments in the development of the law, thereby providing greater leeway for justices' discretion (Sunde 2015,98-101). Thus, goals and motivations are apt to play some role in a justice's decisional behaviour.

Importantly to the comparative utility of our model, the goals or motivations that Baum (1997) and Epstein and Knight (2013) identify and that inform our theoretical perspective are near-universal elements of human behaviour. And as such, they are operational across judicial contexts to varying degrees (see Wijffells & Rhee 2013). The goal of articulating 'good' policy (i.e., policy consistent with the individual justice's attitudinal preferences), for instance, may have greater weight for the justices on Norway's Supreme Court or the German Constitutional Court. In the case of Norway, institutional conditions (e.g., lifetime appointment, discretionary jurisdiction, the power of judicial review) all carve out a large sphere of independence for the justices to sincerely pursue their policy goals when rendering a decision. The German Constitutional Court's duty to safeguard the basic national law results in it preemptively striking down legislation as running afoul of Germany's fundamental law. Moreover, according to De Andrade (2001, 983), the German Court possesses the power to 'compel the legislature to make laws'. In some nations, justices to a country's highest court do not enjoy appointments for terms of good behaviour. Take, for example, Swiss justices. They are elected by the national assembly for six-year terms. Presumably, the motivations of election and retention play a greater role in their decisional calculus. Spanish justices, on the other hand, serve non-renewable nine-year terms, in which case motivations aimed at burnishing a reputation in order to better secure a non-judicial career might come to the fore. As to justices' individual voices, only seven of the 28 member states of the European Union disallow the publication of individual opinions. Interestingly, the traditional distinction of civil and common law systems cut across all 28 states in a way that 'calls into question the validity [of arguments] related to legal cultures and differences in understanding the role of

judges' (Raffaelli 2012,29). Regardless, the key point here is to identify the mix of conditions that affect the relative weight of the various motivations. In this way, a more realistic and transportable model of judicial behaviour might be devised, just as Epstein and Knight (2013) urge.

ACKNOWLEDGMENTS

We are grateful to Marius Svendsen, Kurt-Rune Bergset and Sara Wiest for excellent research assistance, and to Michael R. Tatham for helpful comments. We also thank three anonymous referees who offered incisive comments on a previous version of this article. The article is better for their advice. We presented earlier versions of this article at the Law School and the Department of Comparative Politics, University of Bergen; at the 2014 Norwegian national political science conference, University of Troms0; and at the 72nd Annual Conference of the Midwest Political Science Association in 2014. We are grateful for the comments offered on the article at these and other venues.

1. The literature here is substantial; see, e.g., Pritchett (1948), Schubert (1965), Segal et al. (1995), and Rohde and Spaeth (1976). The definitive treatment is Segal and Spaeth (2002).

2. Approximately 15-20 percent of appealed cases are decided by the Court.

3. If the case addresses a more principled question or the Court expects its decision on the case to challenge earlier rulings, then the case can be heard either in the 11-justice Grand Chamber or in a plenary session consisting of all justices.

4. Please see Online Appendix for details on the variables.

5. The appointment procedure was effectively modified as of 2004.

6. Earlier judge, public prosecutor and interim justice are not tested on the panel level since none are sufficiently frequent to constitute panel majority observations in our data. Age and seniority cannot be used to determine a majority presence on any given panel.

7. On 'error-correcting strategies', see Brenner (1997) and Palmer (1992).

8. See Nicholson and Collins (2008) on the amicus strategy of the US Solicitor General.

9. On the presence of a special relationship between the Solicitor General and the US Supreme Court, see Black and Owens (2013, e.g., 462).

10. Please see the Online Appendix for coding details.

11. Please see the Online Appendix for coding details.

12. See Hox (2010,112, 71 and Chapter 6).

13. This model of judicial behaviour is an improvement over the expanded and intermediate models (see the Online Appendix). Conclusions on model fit and improvements are based on lower AIC/BIC estimates, fewer independent variables from the LR tests, and lower case-level variance proportion by the intra-class correlation (see Appendix Table A1).

14. Appendix Table A2 shows the marginal effects of the independent variables on the probability of voting for the public rights claimant.

REFERENCES

Adolph, C. 2013. Bankers, Bureaucrats and Central Bank Politics: The Myth of Neutrality. New York: Cambridge University Press.

Baum, L. 1997. The Puzzle ofJudicial Behavior. Ann Arbor, MI: University of Michigan Press.

Bentsen, H. L. 2014. 'Dissenser i Norges H0yesterett', Tidsskrift for samfunnsforskning, 55 (1), 3-22.

Beyers, J. 2010. 'Conceptual and Methodological Challenges in the Study of European Socialization', Journal of European Public Policy, 17(6), 909-20.

Black, R. C. & Owens, R. J. 2013. 'A Built-in Advantage: The Office of the Solicitor General and the US Supreme Court', Political Research Quarterly, 66(2), 454-66.

Boyd, C., Epstein, L. & Martin, A. D. 2010. 'Untangling the Causal Effects of Sex on Judging', American Journal of Political Science, 54(2), 389-411.

Brenner, S. 1997. 'Error-Correction on the US Supreme Court: A View from the Clerk's Memos', Social Science Journal, 34(1), 1-9.

Christensen, T., Lsgreid, P. & Zuna, H. R. 2001. 'Profesjonsmangfold, statsmodeller og beslutningsatferd', in Tran0y, B. S. & 0sterud, 0., eds, Den fragmenterte staten. Reformer, makt og styring. Oslo: Gyldendal Akademisk.

De Andrade, G. F. 2001. 'Comparative Constitutional Law: Judicial Review', Journal of Constitutional Law, 3(3), 977-89.

Dyevre, A. 2010. 'Unifying the Field of Comparative Judicial Politics: Towards a General Theory of Judicial Behaviour', European Political Science Review, 2(2), 297-327.

Epstein, L. & Knight, J. 1998. The Choices Justices Make. Washington, DC: Congressional Quarterly Press.

Epstein, L. & Knight, J. 2013. 'Reconsidering Judicial Preferences', Annual Review ofPolitical Science, 16,11-31.

Fagernss, S. O. 2007. 'Vinner stat og kommune for mange saker i H0yesterett?', Advokatbladet (March), 54-5.

Farhang, S. & Wawro, G. 2004. 'Institutional Dynamics on the US Court of Appeals: Minority Representation under Panel Decision Making', Journal of Law, Economics and Organization, 20(2), 299-330.

Fielding, A. 2010. Module 8, Multilevel Modelling in Practice: Research Questions, Data Preparation and Analysis. Available online at: http://www.cmm.bris.ac.uk/lemma/course/ view.php?id=13#m08

George, T. E. & Epstein, L. 1992. 'On the Nature of Supreme Court Decision Making', American Political Science Review, 86(2), 323-37.

Gibson, J. L. 1983. 'From Simplicity to Complexity: The Development of Theory in the Study of Judicial Behavior', Political Behavior, 5(1), 7-49.

Grendstad, G., Shaffer, W. R. & Waltenburg, E. N. 2011.'When Justices Disagree:The Influence of Ideology and Geography on Economic Voting on the Norwegian Supreme Court', Retford, 34(1), 3-22.

Grendstad, G., Shaffer, W. R. & Waltenburg, E. N. 2015. Policy Making in an Independent Judiciary: The Norwegian Supreme Court. Colchester: ECPR Press.

Hanretty, C. 2012. 'Dissent in Iberia:The Ideal Points of Justices on the Spanish and Portuguese Constitutional Tribunals', European Journal of Political Research, 51(5), 671-92.

Hanretty, C. 2013. 'The Decisions and Ideal Points of British Law Lords', British Journal of Political Science, 43(3), 703-16.

Haugland, J. 1986. Dagbok frä Kongens Rad. Oslo: Det Norske Samlaget.

Hönnige, C. 2009. 'The Electoral Connection: How the Pivotal Judge Affects Oppositional Success at European Constitutional Courts', West European Politics, 32(5), 963-84.

Hox, J. J. 2010. Multilevel Analysis: Techniques and Applications. New York: Routledge.

Johnson, S. W., Songer, D. R. & Jiliani, N. A. 2011. 'Judge Gender, Critical Mass and Decision Making in the Appellate Courts of Canada', Journal of Women, Politics and Policy, 32(2), 237-60.

Kastellec, J. P. 2011. 'Panel Composition and Voting on the US Courts of Appeals Over Time', Political Research Quarterly, 64(2), 377-91.

Kj0nstad, A. 1999. 'Er H0yesterett statsvennlig?', Lov og Rett, 2, 97-122.

Lasswell, H. D. 1936. Politics: Who Gets What, When and How. New York: Whittlesey House.

Luo, W. & Kwok, O. M. 2009. 'The Impact of Ignoring a Crossed Factor in Analyzing Cross-classified Data', Multivariate Behavioral Research, 44(2), 182-212.

Lsgreid, P. & Olsen, J. P. 1978. Byräkrati og beslutninger. En studie av norske departement. Bergen: Universitetsforlaget.

Magalhaes, P. C. 2003. The Limits of Judicialization: Legislative Politics and Constitutional Review in the Iberian Democracies. Columbus, OH: Ohio State University Press.

Maltzman, F., Spriggs, J. F. & Wahlbeck, P. J. 2001. Crafting Law on the Supreme Court: The Collegial Game. New York: Cambridge University Press.

McGuire, K. T. 1993. 'Lawyers and the US Supreme Court: The Washington Community and Legal Elites', American Journal of Political Science, 37, 365-90.

Meinke, S. R. & Scott, K. M. 2007. 'Collegial Influence and Judicial Voting Change: The Effect of Membership Change on US Supreme Court Justices', Law and Society Review, 41(4), 909-38.

Michalsen, D. 2006. 'H0yesteretts historie som samfunnshistorie', Nytt Norsk Tidsskrift, 1, 83-95.

Nicholson, C. & Collins, P. M. 2008. 'The Solicitor General's Amicus Curiae Strategies in the Supreme Court', American Politics Research, 36(3), 382-415.

Palmer, J. 1992. 'An Econometric Analysis of the US Supreme Court's Certiorari Decisions', Public Choice, 39(3), 387-98.

Posner, R. A. 2008. How Judges Think. Cambridge, MA: Harvard University Press.

Pritchett, C. H. 1948. The Roosevelt Court: A Study in Judicial Politics and Values, 1937-1947. New York: Macmillan.

Raffaelli, R. 2012. Dissenting Opinions in the Supreme Courts of the Member States. Brussels: European Parliament.

Rohde, D. W. & Spaeth, H. J. 1976. Supreme Court Decision Making. New York: W.H. Freeman.

Rokkan, S. 1967. 'Geography, Religion and Social Class: Crosscutting Cleavages in Norwegian Politics', in Lipset, S. M. & Rokkan, S., eds, Party Systems and Voter Alignments: Cross-national Perspectives. New York: Free Press.

Schei, T. 2010. 'Arbeidet med domsskriving i H0yesterett' in Bergli, W. K. et al., eds, Domsn0kkel. Privatrettslige emner. Oslo: Gyldendal Akademisk.

Schubert, G. 1965. The Judicial Mind: The Attitudes and Ideologies of Supreme Court Justices, 1946-1963. Evanston, IL: Northwestern University Press.

Segal, J. A., Epstein, L., Cameron, C. & Spaeth, H. J. 1995. 'Ideological Values and the Votes of US Supreme Court Justices Revisited', Journal of Politics, 57, 812-23.

Segal, J. A. & Spaeth, H. J. 2002. The Supreme Court and the Attitudinal Model Revisited. New York: Cambridge University Press.

Skiple, J. K. 2015. 'Ei konkret vurdering i kvart enkelt tilfelle. Ein fleirnivaanalyse av 0konomisk stemmegiving i H0gsterett i tidsrommet, 1991-2011', Norsk Statsvitenskaplig Tidsskrift, 31(4), 278-304.

Smith, C. 1975. 'Domstolene og rettsutviklingen', Lov og Rett, 292-319.

Smith, E. 1993. H0yesterett og folkestyret. Pr0vingsretten overfor lover. Oslo: Universitetsforlaget.

Sunde, J. 0. 2012a. 'Andre premiss og anna resultat - refleksjonar kring politikk, H0gsterett og dissensar', Tidsskrift for Rettsvitenskap, 125(1-2), 168-204.

Sunde, J. 0. 2012b. 'Dissenting Votes in the Norwegian Supreme Court, 1965-2009: A Legal Cultural Analysis', Rechtskultur, 1, 59-73.

Sunde, J. 0. 2015. H0gsteretts historie, 1965-2015. Bergen: Fagbokforlaget.

Sunstein, C. R., Schkade, D., Ellman, L. M. & Sawicki, A. 2006. Are Judges Political? An Empirical Analysis of the Federal Judiciary. Washington, DC: Brookings Institution Press.

Tate, C. N. 1981. 'Personal Attribute Models of the Voting Behavior of US Supreme Court Justices: Liberalism in Civil Liberties and Economics Cases, 1946-1978', American Political Science Review, 75(2), 355-67.

Trondal, J. 2007. 'Is the European Commission a "Hothouse" for Supranationalism? Exploring Actor-Level Supranationalism', Journal of Common Market Studies, 45(5), 1111-33.

Voeten, E. 2008. 'The Impartiality of International Judges: Evidence from the European Court of Human Rights', American Political Science Review, 102(4), 417-33.

Volcansek, M. L., ed. 1993. Judicial Politics and Policy-making in Western Europe. London: Frank Cass.

Von Eyben, W. 1956. 'Judicial Law Making in Scandinavia', American Journal of Comparative Law, 5, 112-15.

Wijffells, A. & Rhee C. H. van, eds. 2013. The European Supreme Courts: A Portrait through History. London: Third Millennium Information.

Willoch, K. 2002. Myter og virkelighet. Om begivenheter frem til väre dager med utgangspunkt i perioden, 1965-1981. Oslo: Cappelen.

Supporting Information

Additional Supporting Information may be found in the online version of this article at the publisher's web-site:

Online Appendix

Table A1: Three intermediate justice (J), panel (P) and case (C) models of economic voting on the supreme court 1963-2012

Table A2. Marginal effects of the independent variables on the probability of voting for the public rights claimant