American Law & Economics Association Annual Meetings

Transcription

1 American Law & Economics Association Annual Meetings Year 2005 Paper 9 Does Ideology (and Gender) Matter for Appellate Decision-Making in Canada: A Study of the Ontario Court for Appeal from Moin A. Yahya University of Alberta This working paper site is hosted by The Berkeley Electronic Press (bepress) and may not be commercially reproduced without the publisher s permission. Copyright c 2005 by the author.

2 Does Ideology (and Gender) Matter for Appellate Decision-Making in Canada: A Study of the Ontario Court for Appeal from James Stribopoulos & Moin A. Yahya 1 Faculty of Law University of Alberta Edmonton, AB T6G 2H5 Canada myahya@law.ualberta.ca Draft 1 January 31, 2005 Extremely preliminary. The results are still being computed. Please do not cite or quote without the authors permission. 1 We would like to thank Laurel Lui our capable research assistant who collected the data. We benefited from a Borden Ladner Gervais fellowship to start the research. We would like to thank David Ryan, Janis Sarra, participants at the University of Alberta Faculty of Law workshop, and participants at the University of Alberta economics department for their helpful comments. All errors are ours. 1 Hosted by The Berkeley Electronic Press

3 Does Ideology (and Gender) Matter for Appellate Decision-Making in Canada: A Study of the Ontario Court for Appeal from James Stribopoulos & Moin A. Yahya Faculty of Law University of Alberta Edmonton, AB T6G 2H5 Canada myahya@law.ualberta.ca Abstract In his path breaking work Sunstein (2004) identified ideological trends in voting amongst the United States federal appellate judges for certain categories of cases as well as certain behavior by the judges depending on the ideology of their co-panellists. That trend did not surprise many in the United States where ideological battles over the confirmation of judges to the federal judiciary have become the norm. In Canada, in contrast, the dominant thought among academics and the bar is that ideology, or more precisely the party of appointment, does not matter. In this study, we examine the votes case by the judges on one Canadian appellate court, namely the Ontario Court for Appeal, which adjudicates disputes involving over one third of the population over a fourteen year period. We recorded every reported case from 1990 to We coded each case by the type of case such as criminal law, constitutional law, or private law. We then recorded the votes of each judge on the panel. The party of appointment was recorded for each judge (Liberal or Conservative) as well as the judge s gender. Interesting data patterns emerge. The rate of unanimity is very high at an average of 95%. It varies from a low of 73% in sexual assault cases to a high of 99% in narcotics cases. There is heavy use of the per curiam judgments where they are used one third of the time and to affirm decisions below. We find that party of appointment as well as gender is statistically significant in explaining the outcome of the voting for some categories with gender explaining more than the party of appointment. Given that we had over 12,000 observations and just over sixty judges, we conducted the same tests to see if a specific judge has an ideological bias. We found that certain judges voted overwhelmingly differently from their colleagues on certain categories. Given the high unanimity rate, we also split panels up into two groups: one group had a specific judge on the panel, while the other did not. We found that certain judges presence had a statistically significant impact on whether the panel voted one way or another. These leaders had specific issues for which they led the panel to vote one way. The implications of this study give pause for future research. The first thought is that if ideology matters in both Canada and the United States and in similar categories, then does a theory of judicial behavior exist? What do judges maximize and what are their budgetary constraints that lead them to the outcomes we observe. What policy implications do these studies lead to? 2

4 Should we have more mixed panels by ideology and gender? Perhaps Canada would benefit from judicial selection hearings that the United States currently has. These and many more future research questions are addressed in the paper. 3 Hosted by The Berkeley Electronic Press

5 Does Ideology Matter for Appellate Decision-Making in Canada: A Study of the Ontario Court for Appeal from Introduction James Stribopoulos & Moin A. Yahya In this article we examine whether ideology and gender of the judge matters for the outcome of an appellate case. We follow Sunstein s (2004) study in the United States and use a similar methodology for analyzing the Canadian Appellate Court, the Ontario Court for Appeal during the period. We also examine whether certain judges are leaders on the court in that their presence influences the outcome of a case. We find that certain judges do convince their colleagues to vote contrary to how they would have voted if the judge was absent from the panel. This is true for certain categories. Background The rise of empiricism within law is a relatively recent development. A short while ago, the dominant view was that the law reports alone provided a representative account of what was taking place on the ground in the trial courts (see e.g. Posner (1972)). Although this assumption had long been viewed with much skepticism (Frank (1947), Llewellyn (1960)), it was not until the groundbreaking work of Priest & Klein (1984) that empirical legal research conclusively proved its utility. By examining actual court records, the authors identified a number of recurring variables that created a selection bias in those cases that went to trial -- variables that were unlikely to be present in cases that settled. Of course, this meant that those legal disputes that percolated up on appeal did not provide a representative sampling of all cases. And, with this, empiricism began its ascent within contemporary legal scholarship. 4

6 Although the decisions of appellate courts are no longer considered representative of all the cases in the legal system, they are nonetheless of critical importance. First, appellate courts are usually a litigant s only hope for obtaining relief from trial errors. (Fewer than 1.5% of the cases decided by provincial appellate courts will go on to receive a full hearing before the Supreme Court of Canada (Greene et al. (1998)). Second, these courts also play a central role in shaping the doctrine that influences and controls how cases are dealt with in the trial courts. Therefore, it is not surprising that empirical research investigating appellate courts is on the rise. There have been a number of recent studies in the United States employing quantitative methods to examine various aspects of decision-making by appellate courts. A recent study by Sunstein et al. (2004) examines the possible influence of ideology on the voting patterns of judges on the Federal Court of Appeals. This study uses party of appointment as a proxy for ideology. For an eight year period, it examines the votes of Democratic and Republican appointees with respect to certain categories of reported cases that seem most likely to reveal ideological divisions. (For example, abortion and capital punishment were two of the categories considered). This study finds that in a number of areas the ideology of the judges matters greatly. For example, overall, panels composed entirely of Democratic appointees favour a liberal outcome 61% of the time, whereas panels composed exclusively of Republican appointees prefer a liberal result only 34% of the time. In addition, there are noteworthy results when outcomes involving mixed panels are considered. In short, when a panel contains both Democratic and Republican appointees, the effect of ideology on outcome is significantly dampened. Another study by Ruger et al. (2004) focusses on decision-making by the United States Supreme Court. By examining only a handful of characteristics in cases decided between Hosted by The Berkeley Electronic Press

7 and 2002, the authors are able to formulate a model that enabled them to predict the outcome in 75% of the cases decided by the Court in (In contrast, a panel of legal experts was only able to correctly forecast the outcome in 59.1% of cases.) An even more recent study focuses upon the influence of ideology in the unreported decisions of the Ninth Circuit Federal Court of Appeals (Law (2004)). Apparently, over 80% of the Federal Court of Appeals decisions are unreported judgments that, depending upon the circuit, are either considered to have little or no precedential value. (The same is true in some Canadian appellate courts. For example, the Court of Appeal for Ontario disposes of a great many cases through unpublished endorsements, which due to their brevity are often seen as having little precedential value.) The study finds that for certain judges voting and publication are strategically intertwined. In short, some judges are prepared to acquiesce to decisions that run contrary to their ideological preferences if the case remains unreported, but can be driven to dissent if the majority insists on publication. To date, most of the empirical research on Canadian appellate courts focuses on the Supreme Court of Canada. A great many studies explore various aspects of Supreme Court decision-making. Only quite recently have scholars begun to focus their attention on the provincial appellate courts. It is McCormick a political scientist who pioneered this new and valuable area of empirical research. His early work focusses primarily on the Alberta and Manitoba courts of appeal. With respect to Alberta, he considers its treatment of sentence appeals, in light of a number of different variables (offence, origin of case, panel composition, extent of any modification to sentence) (McCormick (1993a)). Similarly, he also looks at how it deals with criminal conviction appeals (McCormick (1993b)). This last study provides statistics on the 6

8 origins of the cases appealed, the overall success rate, as well as the success rate where the appellant is the accused or the Crown. Statistics as to the processing speed of appeals are also reported. In addition, appeal success rates are given for five general categories of offences. These results are not linked back to the composition of the judges on the panel presiding, although a distinction is made between regular panels and those containing ad hoc judges. More recently, Billingsley & Elman (2001) examine all of the Alberta Court of Appeal s public law decisions since the Charter was enacted that are subsequently overturned by the Supreme Court. They conclude that some fundamental philosophical differences divide the two courts in many of these cases. McCormick s studies of the Manitoba Court of Appeal are also quite extensive. He analyses all of the Manitoba Court of Appeal cases that were reviewed by the Supreme Court between 1906 and This study attempts to identify recurring patterns or variables amongst these cases. (McCormick & Maisey (1992b)) In addition, he looks at the court s caseload and output over a four year period (McCormick (1990), McCormick (1991b), McCormick (1992a) & McCormick (1993c)). With respect to output, these studies report upon a number of relevant features of the court s decisions. First, the success rates of appellants before the court generally, and then with respect to specific types of cases (categories include: sentence, criminal, family, private law, public law, financial, and reference). Second, the voting patterns of specific judges are also considered generally. That is, the statistical rate at which each judge votes to dismiss or allow appeals. And, more specifically, the study also considers how each judge votes in criminal appeals. So, for example, the statistical rate at which each judge favours the Crown as opposed to the accused in criminal appeals is reported. (Cases within the criminal category are not broken down further based on offence type or the fact that a Charter claim may have been involved.) 7 Hosted by The Berkeley Electronic Press

9 These studies also report on a number of other aspects relating to decision-making at the Manitoba Court of Appeal, including: the rate at which individual judges dissent from the majority; apparent alliances between certain judges based on frequency of pairings in nonunanimous decisions; the source of precedents cited by the court in its judgments; and the rate at which precedents from various courts are cited in specific categories of cases. Finally, the most recent study of the Manitoba Court of Appeal also includes analysis of an additional variable, success rates based on the type of litigant (i.e. Crown, Big Business, Fed/Prov. Gov t, Business, Municipal Gov t, Individuals, Other Litigants) (McCormick (1993)). So, for example, the study reveals that in cases decided between Big Business succeeded before the court 65.7% of the time, whereas individuals only succeeded 38.5% of the time. There are also analogous studies of the Saskatchewan Court of Appeal (Newman (2001), Newman (2002), Newman (2004)). In each, Newman examines the output of the Saskatchewan Court of Appeal, both from a quantitative and qualitative standpoint. His studies include statistics on the types of cases decided, the number of decisions rendered by each judge (and the form of those decisions, ie. written or oral), as well as the voting record of each judge generally, and more specifically in both civil and criminal cases. Finally, the voting record of the judges is broken down further, depending upon whether the Crown or an accused was appealing, and whether the appeal was a conviction or sentence appeal. (Beyond these general categories, however, none of these studies track the voting of individual judges with respect to more discrete legal issues within each category.) McCormick and others also studied the Court of Appeal for Ontario, focusing on the processing of appeals by that court. (Baar et al. (1992)) This group collaborated again, publishing the first comprehensive study on appellate courts in Canada (Greene et al. (1998)). 8

10 The study employs both qualitative and quantitative methods. For example, 80% of the judges sitting on Canadian appellate courts are interviewed, and a representative sample of 6000 cases from these courts is analyzed. The empirical analysis of the cases yields statistical insights into various aspects of decision-making by appellate courts across the country, including: the stage of the process at which criminal appeals are resolved (before or after hearing); the rate at which judgments are reserved on criminal and civil cases; the attributes of decisions (details about length and citations); the historic rates of dissenting opinions and large panels presiding over appeals; the speed at which appeals are processed and heard; the length of hearings; the amount of time that elapses between hearing and judgment; and the rate at which substantive criminal appeals are allowed and dismissed. Our study creates a comprehensive database of the Ontario Court for Appeal s reorted decisions over a fourteen-year period. We undertake in Canada the kind of empirical research into appellate court decision-making that has been on the rise in the United States as of late (see Sunstein et al. (2004), Ruger et al. (2004), Law (2004)). This research would make a valuable contribution to Canadian political and legal discourse. Today, opinions about appellate courts and judges are largely impressionistic, the result of anecdotal observations by lawyers and legal scholars. This study would provide a means of empirically confirming or debunking such claims. So, for example, in the public debate short-lived as it was regarding the most recent appointments to the Supreme Court, it was alleged by some that Justice Rosalie Abella was an activist judge (Matas (2004)). Our research would provide a means of testing such claims, by enabling the decision-making of individual judges to be empirically evaluated. Given that Supreme Court judges are almost 9 Hosted by The Berkeley Electronic Press

11 always drawn from the provincial appellate courts, and the current government s commitment to a more transparent appointments process, the importance of such research cannot be overstated. Methodology We collected data from the reported decisions of the Court of Appeal for Ontario between 1990 and 2003 (so far, we have only electronically inputted ten years of data ). We categorized each case into the broad categories of Criminal Law, Public Law, Commercial Law, Family Law, Employment Law, Labor Law, and Other. For Criminal Law (where the actors were the Crown and the accused ), we created the sub-categories of Crimes Against the Person (ie. Assault, robbery), Property Offences (ie. Theft, fraud), Public Morals (ie. Gambling, prostitution), Narcotics (possession, sale or importing), Sexual Assault, Charter Case Specific Remedy (seeking stay or evidentiary exclusion), Charter Invalidation (seeking declaration that law unconstitutional), and Charter Combined (seeking declaration and exclusion). For Public Law, the sub-categories were Non-Criminal Charter (where we identified the plaintiff and defendant as: individual, corporation, Crown, or other state actor ), Constitutional Division of Powers (plaintiff and defendant are: individual, corporation, provincial government, federal government or other state actor ), Human Rights (plaintiff and defendant are: individual, corporation, Crown or human rights commission ), Administrative Law (plaintiff and defendant are: individual, corporation, Crown, or administrative body ), Environmental Law (plaintiff and defendant are: individual, corporation, Crown or other state actor, Aboriginal Charter Claims (the parties are: individual, band, provincial government, federal government or other state actor ), and 10

12 Aboriginal Other Claims (ie. Land claims) (plaintiff and defendant are: individual, band, provincial government, federal government or other state actor ). For Commercial Law, the sub-categories are Contracts (plaintiff and defendant are: individual, corporation, Crown, or other state actor ), Corporate Law: Oppression Remedy (plaintiff and defendant are minority shareholder(s) and majority shareholder(s), Corporate Law All Other, Insurance Law (parties are corporation or individual ), and Professional responsibility (plaintiff and defendant are client or lawyer ). For Family Law (plaintiff and defendant are husband/father or wife/mother ), the subcategories are Child Custody, Child Support, Spousal Support, and Other. For Employment and Labour Law, the sub-categories are Wrongful Dismissal (plaintiff and defendant are employee and employer ), and Labour (plaintiff and defendant are union, employee or management ). These categories proved too detailed and resulted in too few observations. So we collapsed the categories into slightly broader categories. These are as follows. Aboriginal: AboriginalCharter + AboriginalOther. AdminLaw: AdminLaw + Taxation. ConLaw: Constitutional. Private: Contracts + CorpOppression + CorpOther + Insurance + Torts + Negligence. CrimChrtr: CrimCharterCSR + CrimCharterCombined + CrimCharterInvalid. CrimBL: CrimDomestic + CrimNarcotics + CrimOffencesPerson + CrimOther + CrimProperty + CrimPublicMorals + CrimSexual. Public: CrimReg + Environmental. Family: FamilyChildCustody + FamilyChildSupport + FamilyOther + FamilySpousalSupport. Labor: Labour + WrongDismissal. HRT: NonCrimCharter + HumanRights. CrimGender: CrimDomestic + CrimSexual. 11 Hosted by The Berkeley Electronic Press

13 Results: Descriptive Statistics Between 1994 and 2003, 2 there were a total of 4092 reported cases, 3 for which panels of mostly three judges sat. 4 This generated 11,810 observations. Table 1 contains the number of reported cases by year. Table 2 reports the number of cases by the narrow categories, while table 2a reports them by the broader classifications. The category that is dealt with the most is criminal black letter appeals followed by private disputes. Table 3 reports the dispositions of the cases below, while table 4 reports the outcomes of the cases on appeal. Slightly less than 60% of the appeals result in an affirmation of the decision below, while just over 40% result in overturning the decision below. 5 Table 5 reports the unanimity rate for all categories. The rate increases to 99% in narcotics cases where an accused was convicted below, and similarly, it is 99% in criminal Charter cases where a case specific remedy (such as suppression of the evidence) was sought and denied by the trial court (of course, many of these are also drug cases). In contrast, it decreases to 86% in cases where the accused was acquitted of sexual assault at trial, 6 and it decreases even further to 73% in family law cases where a parent unsuccessfully sought custody of a child at trial. Needless to say these are cases where there is little consensus on what the law is or how it should applied among broad spectrums of society. Hence, it is not surprising to see lower unanimity rates in such categories. The use of per curiam judgments is quite pervasive. Such judgments were used approximately in one third of all cases as table 6 shows. Furthermore, per curiam decisions were 2 While the abstract and introduction refers to our sample being from , we have not yet coded or analyzed the years Hence, the results in this version cover about ten years. We hope to have the rest of the years in the analysis by the end of April We address the question of unreported cases later in the discussion. 4 Sometimes panels of five judges sit for what is the equivalent of en banc rehearings. Sometimes one of the five panelists becomes unavailable, and some decisions are rendered with four judges. 5 Discuss relevance to Priest Klein model. 6 In Canada, the state (or crown as it is called) can appeal acquittals as the prohibition against double jeopardy does not apply in that situation in Canada. 12

14 used almost exclusively to affirm decisions below. They were used in a statistically significant way to affirm cases involving contracts, charter cases where a remedy was requested (such as suppression of the evidence), criminal domestic violence cases, criminal narcotics cases, crimes of violence, and professional responsibility. Finally, tables 7-10 show the composition of the panels by gender and party. The period is rich with judges of both parties, since the Liberal Party of Canada took power in late 1993 after the Progressive Conservative Party of Canada had been in power since We have panels with roughly half the judges voting from one of the two parties. Additionally, female votes account for about 25% of the votes. Table 11 shows the results of the first Logistic regression (Logit) that we ran. The equation is as follows: P( y i ) = i x + { i i 123 yi = 1 Affirm case category interacted with Decision below yi = 0 Overturn The left hand side variable is the judge s vote: affirm or overturn. The right hand side is a set of interactive dummies comprised of the case category interacted with the decision below (no intercept was used). The results in table 11 show that Administrative law cases where the case was dismissed below are affirmed in a statistically significant manner (63%), as are constitutional law challenges brought below that are dismissed (91%). This seems to suggest that there is a great deference to the administrative state, since decisions of administrative agencies must be challenged in the trial court first. If the trial judges dismisses the challenge, i.e. defers to the administrative agency, the appeal court is more likely to affirm the dismissal. The criminal law black letter appeals are also significant. Where an acquittal was had below, there is a strong chance of overturning the verdict (74%) perhaps reflecting a selection 13 Hosted by The Berkeley Electronic Press

15 bias by the crown choosing to appeal only the most egregious acquittals. Convictions are affirmed 60% of the time in a significant fashion. Finally and notably the disposition of private disputes is affirmed in statistically significant manner. Results: Does Ideology Matter? We conducted logit regressions with gender only and political affiliation only as a independent variables (interacted with the case category and disposition below) and found that gender and political affiliation did matter for certain categories. The results of the regressions with the judge s party of appointment as a variable are in table 11b. Although conservatives and liberals generally voted the same way on a variety of issues, what distinguishes them is a slight difference on the probability of affirming or overturning the disposition below. For example, liberals were less likely to affirm a conviction than a conservative judge, but only by a slight amount. Table 11a shows the results by the judge s gender. An interesting result for example is in crimes of gender, i.e. sexual assault and domestic violence. Male judges increased sentences when the crown appealed, and affirmed sentences when the defendants appealed, as did the female judges. The female judges were 20% more likely than their male counterparts, however, to increase sentences of defendants when the crown appealed. Furthermore, female judges were statistically significantly likely to overturn an acquittal of an accused charged with a gender specific crime. On the other hand, both male and female judges were statistically likely to affirm convictions and overturn acquittals on an equally likely basis for non-gender specific crimes (with females slightly less likely to affirm convictions but slightly more likely to overturn acquittals). Gender seems to explain more of the variation than party of appointment. 14

16 [We have not yet interacted by party and gender of individual judge] Then we conducted a logit regression where the left hand-side was the judge s vote, and the right hand side included interacted dummies. The dummies interacted were the party mix of the panel, the gender mix of the panel, the case category, and the disposition below. Table 12 conveys these results. Under panel gender, FF refers to an all female panel, FM refers to a mix gender panel, and MM refers to an all male panel. 7 For panel party, CC refers to al conservative panel, LL refers to an all liberal panel, and LC refers to a mixed panel. Notable results are as follows. In the area of criminal black letter convictions, mixed gender all conservative panels are statistically significant in affirming a conviction as are mixed gender mixed party panels, mixed gender all liberal panels, all male all conservative panels, and all male mixed party panels. The difference is the intensity or likelihood of convicting. Calculating the odds ratios using the estimated coefficients yields the probability of affirmation by a mixed gender all conservative panel as 62%, by a mixed gender mixed party panel as 60%, mixed gender all liberal panel as 58%, all male all conservative panels as 70%, and all male mixed party panels as 62%. Clearly gender and party matters. All male and all conservative panels have a 70% likelihood chance of affirming a criminal conviction, while mixed gender all liberal panels have only a 58% likelihood. Even all conservative panels with at least one female lowers the likelihood of conviction down to 62%. An all male mixed party panel also has a 62% chance of conviction, which suggests that the male conservative judges are the outliers, while ideology and gender does not matter for the rest of the panels. Similarly in the area of gender specific criminal cases, party and gender mattered. Mixed gender all conservative panels affirm conviction and overturn acquittals, mixed party mixed 7 Since the panels are mostly unanimous, breaking down by whether there are two females and one male or two males and one female would result in multicollinearity. 15 Hosted by The Berkeley Electronic Press

17 gender panels overturn acquittals, all male mixed party affirm convictions, and all male all liberal panels actually overturn such convictions. Again the intensity of the voting changes by composition. All conservative mixed gender panels are 64% likely to affirm a conviction and 75% likely to overturn an acquittal, while mixed party and mixed gender panels overturn acquittals 88% of the time suggesting that a liberal on a mixed gender panel can add to the likelihood of overturning an acquittal by 13%. All male mixed party only affirm convictions 56% of the time, while all male all liberal panels overturn convictions 70% of the time. Similar results can be seen for other categories. Other Results: Do Individual judges matter? We re-did the regressions with individual judges interacted with the category and disposition below. We found that certain judges do stand out. [This data is massive (over fifty judges in the sample) and while the results are available electronically, we have not compiled them yet for this paper] These judges stand out on certain categories. We were surprised that certain well known judges did not stand out in high-profile areas. This can be explained by the high unanimity rate in the voting. Given that there is a high unanimity rate, it would seem then that a strong-willed judge might be able to influence the outcome of a case. We tested this hypothesis by looking at panels where a judge was present and panels where a judge was absent to see if that made a difference. We found that for certain judges, their presence did make a difference for certain categories. For example, a certain judge [we may reveal the names in the final draft as these are preliminary results] who has a reputation for being an activist did not have any individual votes that were different from the other judges. But when we studied whether this judge is a leader, we found 16

18 some interesting results that are in Table 13. We regressed the outcomes of the cases on an interacted dummy of the category, disposition below, and whether the judge was present or absent on the panel. In the area of criminal law, the absence of the judge meant that acquittals were more likely to be reversed and convictions affirmed. The judge s presence meant that the conviction was affirmed but 4% less likely. Similarly, gender crime convictions were more likely to be affirmed when the judge was present then when the judge was absent. This suggests that certain judges can lead by their presence or absence. We intend to do this exercise for all judges in our sample. Preliminary results indicate that some judges are true leaders, while others do not exhibit statistical influence on their colleagues. Future Work We hope to understand the process of judicial decision-making more broadly than the Ontario Court for Appeal. How do these results compare with Sunstein s for example? What implications do these results have for judicial selection generally? These are some of the questions we wish to address in this paper and future research. 17 Hosted by The Berkeley Electronic Press

19 Tables Table 1: Number of cases by year. Number of Year cases Table 2: the number of cases by narrow classification. Number Category of Cases Percent AboriginalCharter AboriginalOther AdminLaw Constitutional Contracts CorpOppression CorpOther CrimCharterCSR CrimCharterCombined CrimCharterInvalid CrimDomestic CrimNarcotics CrimOffencesPerson CrimOther CrimProperty CrimPublicMorals CrimReg CrimSexual Environmental FamilyChildCustody FamilyChildSupport FamilyOther FamilySpousalSupport

20 HumanRights Insurance Labour Negligence NonCrimCharter Other ProfResponsibility Taxation Torts WrongDismissal Table 2a: the number of cases by broad categories. Number Case Type of Cases Percent Aboriginal AdminLaw ConLaw CrimBL CrimChrtr CrimGN Family HRT Labor Other PR private public Table 3: Disposition of cases at trial. Number Decision Below of Cases Percent Magnitude acquittal allowed conviction dismissal plea stay Hosted by The Berkeley Electronic Press

21 Table 4: Decision on Appeal Decision on Appeal Number of Cases Percent affirm overturn stay Table 5: Unanimity Rate Decision Number of Cases Percent Dissent Unanimous Table 6: Use of Per Curiam Per Curiam Frequency Percent No Yes Table 7: Presence of Female Judge At least on female on panel? Frequency Percent Yes No Table 8: Presence of Conservative appointee At least one Conservative Appointee on Panel? Frequency Percent No Yes Table 9: Gender Composition Gender of Judges Voting Number of Votes Percent F M

22 Table 10: Political Composition Party of Number Appointmentof Votes Percent C L Table 11: Affirmance by Category and disposition below bold reflects statistically significant at 10% level. Average Estimate Std. Err. Wald 2 Affirmancep-value Aboriginal Acquittal Aboriginal Allowed Aboriginal Dismissal AdminLaw Magnitude AdminLaw allowed AdminLaw dismissal % AdminLaw stay ConLaw allowed ConLaw dismissal % CrimBL Magnitude CrimBL acquittal % CrimBL allowed CrimBL conviction % <.0001 CrimBL dismissal CrimBL plea CrimBL stay % CrimChrtr Magnitude CrimChrtr acquittal % CrimChrtr allowed CrimChrtr conviction % <.0001 CrimChrtr dismissal % CrimChrtr stay % CrimGN Magnitude % CrimGN acquittal CrimGN conviction CrimGN plea CrimGN stay Family Magnitude Family allowed % Family dismissal HRT Magnitude Hosted by The Berkeley Electronic Press

23 HRT allowed HRT dismissal % <.0001 Labor Magnitude Labor allowed Labor dismissal Other Magnitude Other allowed Other conviction Other dismissal Other stay PR Magnitude PR allowed PR dismissal % private Magnitude % private allowed % <.0001 private conviction private dismissal % <.0001 public Magnitude public acquittal % public allowed public conviction public dismissal public stay Table 11a: Affirmance by Category, disposition below, and judge s gender bold reflects statistically significant at 10% level. Gender Category Disposition Below Estimate Std. Err. Wald Chi2 p-value F Aboriginal allowed F Aborigina l dismissal F AdminLaw MagDF F AdminLaw allowed F AdminLaw dismissal F AdminLaw stay F ConLaw allowed F ConLaw dismissal F CrimBL MagCR F CrimBL MagDF F CrimBL acquittal F CrimBL allowed F CrimBL conviction <

24 F CrimBL dismissal F CrimBL plea F CrimBL stay F CrimChrtr MagCR F CrimChrtr MagDF F CrimChrtr acquittal F CrimChrtr allowed F CrimChrtr conviction F CrimChrtr dismissal F CrimChrtr stay F CrimGN MagCR F CrimGN MagDF F CrimGN acquittal F CrimGN conviction F CrimGN stay F Family MagHL F Family allowed F Family dismissal F HRT MagCR F HRT MagDF F HRT allowed F HRT dismissal F Labor MagCR F Labor MagDF F Labor allowed F Labor dismissal F Other MagCR F Other MagDF F Other allowed F Other conviction F Other dismissal F PR MagDF F PR allowed F PR dismissal F private Mag-Appless F private Mag- Appmore F private allowed <.0001 F private dismissal <.0001 F public MagCR F public acquittal F public allowed F public conviction F public dismissal Hosted by The Berkeley Electronic Press

25 F public stay M Aboriginal acquittal M Aboriginal allowed M Aboriginal dismissal M AdminLaw MagDF M AdminLaw allowed M AdminLaw dismissal <.0001 M AdminLaw stay M ConLaw allowed M ConLaw dismissal <.0001 M CrimBL MagCR M CrimBL MagDF M CrimBL acquittal <.0001 M CrimBL allowed M CrimBL conviction <.0001 M CrimBL dismissal M CrimBL plea M CrimBL stay M CrimChrtr MagCR M CrimChrtr MagDF M CrimChrtr acquittal M CrimChrtr allowed M CrimChrtr conviction <.0001 M CrimChrtr dismissal <.0001 M CrimChrtr stay M CrimGN MagCR M CrimGN MagDF <.0001 M CrimGN acquittal M CrimGN conviction M CrimGN plea M CrimGN stay M Family MagHL M Family MagWL M Family allowed <.0001 M Family dismissal M HRT MagCR M HRT MagDF M HRT allowed M HRT dismissal <.0001 M Labor MagCR M Labor MagDF M Labor allowed M Labor dismissal

26 M Other MagCR M Other MagDF M Other allowed M Other conviction M Other dismissal <.0001 M Other stay M PR MagCR M PR MagDF M PR allowed M PR dismissal M private Mag-Appless M private Mag- Appmore M private Magnitude M private allowed <.0001 M private conviction M private dismissal <.0001 M public MagCR M public MagDF M public acquittal <.0001 M public allowed M public conviction M public dismissal M public stay Table 11b: Affirmance by Category, disposition below, and judge s party of appointment bold reflects statistically significant at 10% level. Party Category Disposition Below Estimate Std. Err. Wald Chi2 p-value C Aboriginal acquittal C Aboriginal allowed C Aboriginal dismissal C AdminLaw Magnitude C AdminLaw allowed C AdminLaw dismissal C AdminLaw stay C ConLaw allowed C ConLaw dismissal <.0001 C CrimBL Magnitude C CrimBL acquittal <.0001 C CrimBL allowed C CrimBL conviction <.0001 C CrimBL dismissal Hosted by The Berkeley Electronic Press

27 C CrimBL plea C CrimBL stay C CrimChrtr Magnitude C CrimChrtr acquittal C CrimChrtr allowed C CrimChrtr conviction <.0001 C CrimChrtr dismissal <.0001 C CrimChrtr stay C CrimGN Magnitude C CrimGN acquittal C CrimGN conviction C CrimGN plea C CrimGN stay C Family Magnitude C Family allowed C Family dismissal C HRT Magnitude C HRT allowed C HRT dismissal <.0001 C Labor Magnitude C Labor allowed C Labor dismissal C Other Magnitude C Other allowed C Other conviction C Other dismissal <.0001 C Other stay C PR Magnitude C PR allowed C PR dismissal C private Magnitude C private allowed <.0001 C private conviction C private dismissal <.0001 C public Magnitude C public acquittal C public allowed C public conviction C public dismissal C public stay L Aboriginal acquittal L Aboriginal allowed L Aboriginal dismissal

28 L AdminLaw Magnitude L AdminLaw allowed L AdminLaw dismissal L ConLaw allowed L ConLaw dismissal L CrimBL Magnitude L CrimBL acquittal L CrimBL allowed L CrimBL conviction <.0001 L CrimBL dismissal L CrimBL plea L CrimBL stay L CrimChrtr Magnitude L CrimChrtr acquittal L CrimChrtr allowed L CrimChrtr conviction <.0001 L CrimChrtr dismissal L CrimChrtr stay L CrimGN Magnitude L CrimGN acquittal L CrimGN conviction L CrimGN stay L Family Magnitude L Family allowed <.0001 L Family dismissal L HRT Magnitude L HRT allowed L HRT dismissal <.0001 L Labor Magnitude L Labor allowed L Labor dismissal L Other Magnitude L Other allowed L Other conviction L Other dismissal <.0001 L Other stay L PR Magnitude L PR allowed L PR dismissal L private Magnitude L private allowed <.0001 L private conviction L private dismissal < Hosted by The Berkeley Electronic Press

29 L public Magnitude L public acquittal L public allowed L public conviction L public dismissal L public Stay Table 12: Affirmance by Category, disposition below, panel party, and panel gender bold reflects statistically significant at 10% level. Panel-Gender Panel- Party Category Decision Below Estimate Std. Err. Wald p-value FF CC CrimBL Magnitude FF CC CrimBL conviction FF CC CrimChrtr dismissal FF CC CrimChrtr stay FF CC CrimGN conviction FF CC Private allowed FF CC private dismissal FF CC public stay FF LC CrimBL allowed FF LC CrimBL conviction FF LC Other Magnitude FF LC Other allowed FF LC Other dismissal FF LC private allowed FF LL Other allowed FF LL private allowed FF LL private dismissal FM CC AdminLaw allowed FM CC AdminLaw dismissal FM CC AdminLaw stay FM CC ConLaw allowed FM CC ConLaw dismissal FM CC CrimBL Magnitude FM CC CrimBL acquittal FM CC CrimBL allowed FM CC CrimBL conviction FM CC CrimBL dismissal FM CC CrimBL stay FM CC CrimChrtr acquittal FM CC CrimChrtr allowed FM CC CrimChrtr conviction

30 FM CC CrimChrtr dismissal FM CC CrimChrtr stay FM CC CrimGN Magnitude FM CC CrimGN acquittal FM CC CrimGN conviction FM CC CrimGN stay FM CC Family Magnitude FM CC Family allowed FM CC Family dismissal FM CC HRT allowed FM CC HRT dismissal FM CC Labor Magnitude FM CC Labor allowed FM CC Labor dismissal FM CC Other allowed FM CC Other conviction FM CC Other dismissal FM CC PR allowed FM CC PR dismissal FM CC private Magnitude FM CC private allowed FM CC private dismissal FM CC public allowed FM CC public conviction FM LC Aboriginal allowed FM LC Aboriginal dismissal FM LC AdminLaw Magnitude FM LC AdminLaw allowed FM LC AdminLaw dismissal FM LC ConLaw allowed FM LC ConLaw dismissal FM LC CrimBL Magnitude FM LC CrimBL acquittal <.0001 FM LC CrimBL allowed FM LC CrimBL conviction <.0001 FM LC CrimBL dismissal FM LC CrimBL plea FM LC CrimBL stay FM LC CrimChrtr Magnitude FM LC CrimChrtr acquittal FM LC CrimChrtr allowed FM LC CrimChrtr conviction <.0001 FM LC CrimChrtr dismissal Hosted by The Berkeley Electronic Press