Direct Measures of Ideological Difference
The effect of judges’ ideology on consensus on a court is a contentious topic in empirical studies of judicial decision-making. We might reasonably expect some variation across high courts, with ideology mattering more in countries with a politicized judicial system than for courts set in a more apolitical context, but even this is not near the mark. There is no clear consensus on the connection between ideology and agreement even within a given country.
Studies of the uS Supreme Court differ on how ideology interacts with norms of consensus. Wahlbeck, Spriggs, and Maltzman’s study of the Burger Court found that the justices’ ideology does indeed affect rates of consensus, as justices are “more likely to author or join a separate opinion the further they are ideologically from the majority opinion author" However, ideology is generally seen as one factor acting in combination with other considerations. Consider the work of Walker, Epstein, and Dixon (1988), which documents a large jump in dissents on the US Supreme Court. Walker, Epstein, and Dixon find that ideology may at times result in coalitions on the court, but that ideology alone cannot account for the sudden increase in separate judgments during the Stone era. A more compelling story, to them, is that there was a “generalized decline in the norms of cohesion" In contrast, Haynie hypothesizes that courts with greater divisions along party lines will produce more separate opinions, but ultimately finds no relationship between ideology and dissent or concurrence.
More recently, Edelman, Klein, and Lindquist examined whether ideology lies behind both disagreement and agreement. They analyzed the relationship between the degree of disordered voting (that is, voting that is not in line with ideological expectations) and the level of disagreement. They found that for cases where there was a greater predicted level of disorder (non-ideological voting), there were fewer dissents. Although ideology leads to ordered voting and more dissents, there is greater consensus in disordered cases, and so ideological voting is not a factor (or at least not the only factor) leading to agreement.
Differences in attitudes may matter to the dissent rate at the Supreme Court of Canada. Songer, Szmer, and Johnson examined the role of ideology (among other factors) in the rate of dissents on the Court between 1976 and 2005. To measure attitudinal difference, they used the difference in career liberalism scores for the most liberal and the most conservative justices on a particular panel. After controlling for a range of factors including panel size, legal area, workload, and chief justice, they found a positive and statistically significant relationship between attitudinal difference and dissent rates such that a one standard deviation increase in the attitudinal difference increases the probability of a dissent by about 22 percent.
Studies of the Australian High Court, on the other hand, tend not to find that ideology has had a significant effect on consensus. For example, in the Latham era (1935-1950) justices’ ideological compatibility was not significantly related to their likelihood of agreement. On the other hand, between 1904 and 2001, when a conservative government was in power, conservative appointees dissented less often than Labor appointees, and judges who had a political career in the Labor party dissented more often overall.51
One measure of disagreement we can use to examine this issue of ideology and dissent across different courts is the number of opinions with at least one dissent. If we look at our data for 1970 to the early 2000s and measure the size of the ideological differences on a court by looking at the spread of the lifetime liberal voting scores of judges on that court in a given year, Figure 8-4 shows the predicted dissent rate across a range of such differences.52 The horizontal axis shows the potential spread of liberal voting rates of judges on a court with the spread measured by the standard deviation of the lifetime liberal voting scores for the judges on the court in a given year. It encompasses the range of spreads on these courts in the period 1970 to the early 2000s. The Supreme Court of Canada in fact tended to have a lower spread of ideology (with a mean over the period of 0.051) and the united States a higher spread (with a mean of .148). The vertical axis is what we would predict the percent of opinions with at least one dissent would be at each of the possible differences in liberal voting rates.
The change in the differences in the ideological spread as measured by lifetime liberal voting rates does appear related to the dissent rate, but the impact varies considerably across courts. The uS and Australian predicted rate of dissents is high across the spread of liberal votes rates, though the Australian High Court dissent rate is much more responsive to increases in differences in liberal voting rates on the Court. The dissent rate on the Supreme Court of Canada is also highly responsive to changes in the composition of the Court. The uK dissent rate is low and not as responsive to changes on the Court as the Court in Australia and Canada.
the relative conservatism of the judges. Glendon Schubert, “Opinion Agreement among High Court Justices in Australia” (1968) 4:1 Australian and New Zealand Journal of Sociology 2 at 15. In a broader study covering 1903 to 1975 Smyth finds that the commonly held hypothesis that Labor party judges dissent more often is only very weakly borne out in some tests and not in others. Russell Smyth, “Explaining Historical Dissent Rates in the High Court of Australia” (2003) 41:2 Journal of Commonwealth and Comparative Politics 83 at 107 [Smyth, “explaining Historical Dissent”]. Smyth notes that by stopping in 1975, the paper misses several Labor appointees known to be above average dissenters (at 109).
- 51. Russell Smyth, “The Role of Attitudinal, Institutional and environmental Factors in explaining Variations in Dissent Rates on the High Court of Australia” (2005) 40:4 Australian Journal of Political Science) 519 at 535 [Smyth, “Role of Attitudinal”].
- 52. See Appendix 3 for a discussion of the variables used and the results of the related regressions. note that the marginal values for idsd was significant at the .05 level for each country.
Figure 8-4 Predicted percentage of opinions with at least one dissent based on the difference in the liberal voting rate on each court, 1970s to early 2000s. The difference in the liberal voting rate is the mean of the annual standard deviations of the lifetime liberal voting rates of judges on the court over the period. The predictions were found controlling for time and workload (total cases heard in the year). Canada and Australia were the most responsive to alterations in the composition of the court; the united States and the uK were much less so.
We can also measure how likely judges are to dissent by looking at the number of dissenting opinions per year—that is, not the number of cases per year that have dissents, but the number of times in a year that judges decide to dissent. To take account of differences in the caseload per year, we need to look at the fraction of dissenting opinions, which we find by dividing the number of dissenting opinions in a year by the number of decisions that year. We saw in Chapter 1 (Figure 1-3) that there was a large difference across countries in terms of the fraction dissenting, from a high in the united States of almost 1.5 dissenting opinions for every decision down to almost zero for India. Our question now is—does the size of the difference in ideology on a court in a given year correlate with the differences in the fraction of dissenting opinions? It turns out it does in Canada, the uK, and the united States but does not in India and Australia. Assume again that we can measure the size of the ideological differences across courts by looking at the spread of the lifetime liberal voting scores
Figure 8-5 Predicted percentage of fraction of opinions that are dissenting based on the difference in the liberal voting rate on each court, 1970s to early 2000s.54 The difference in the liberal voting rate is the mean of the annual standard deviations of the lifetime liberal voting rates of judges on the court over the period. The predictions were found controlling for time and workload (total cases heard in the year). Justices on the Supreme Court of Canada appear most sensitive to changes in the composition of the Court.
of judges on the court in that year.54   Figure 8-5 shows the connection between this spread of liberal voting rates and the fraction of dissenting opinions on a court. We did not include India and Australia as the relationship was not statistically significant in these countries.
As in Figure 8-4, the horizontal axis shows the potential spread of liberal voting rates of judges on a court (with the spread measured by the standard deviation of the lifetime liberal voting scores for the judges on the court in a given year). The vertical axis is what we would predict the fraction of dissenting opinions would be at each of the possible differences in liberal voting rates. The figure shows that the uS Supreme Court has a fairly high baseline dissenting rate. Even if the justices were fairly similar on the Court (to the left end of the horizontal axis), the justices would be relatively likely to dissent. However, this high dissenting rate is not particularly sensitive to increases in how diverse the justices on the Court are given the fraction dissenting rises relatively slowly as the voting tendencies on the Court become more spread out. The uK House of Lords, on the other hand, has a low base rate level of dissent, with justices relatively infrequently writing dissenting opinions. Interestingly, the uK House of Lords is also relatively insensitive to changes in the range of voting of its members. It has a low rate of dissenting even on a highly diverse court. The fraction dissenting on each of these courts could potentially be understood through norms. The uS Supreme Court has a norm of dissenting and the uK House of Lords of not dissenting, both of which are fairly insensitive to the membership of the Court.
The Supreme Court of Canada, on the other hand, appears very sensitive to the voting behavior of its members. Like the uK House of Lords, it has a fairly low baseline of dissenting behavior, with a low fraction of dissenting opinions when the Court is fairly similar. However, as the voting behavior of the justices on the Court diverges (we move to the right on the figure), the justices on the Court are much more likely to dissent. The difference from the uK and the united States cannot be explained solely through the use of panels, because although the united States hears all cases en banc, the Canadian and uK courts hear decisions in panels. So in Canada the story appears slightly more complex than a possible consistent norms story.
We could also look at how individual judges vote to try to find the effect of differences in policy preferences. Figure 8-6 examines whether differences of a judge from her fellow panelists influences her probability of dissenting. Moving from left to right along the horizontal axis represents moving from the judge having a similar lifetime liberal voting rate to the average of the other panelists to the judge having a progressively higher or lower lifetime liberal voting than the other panelists (that is, it is absolute value of the difference). The vertical axis is the probability of the judge dissenting. There are two key facts to note about the Figure. First, a judge in the united States has the highest
Figure 8-6 Influence on the probability of a judge dissenting of the difference of the judge’s lifetime liberal voting rate from the average lifetime liberal voting rate of the other panelists (percent). At the left hand side the judge’s lifetime liberal voting rate is the same as the average of the other panelists. At the right hand side, the judge’s lifetime liberal voting rate is 30 percent higher or lower than the average of the other panelists. The United States and Australia have the highest base rate of dissenting (the probability of a justice dissenting even if he is at the average of the other panelists), and Canada and the UK have the lowest. Justices in Canada and Australia are the most responsive to differences from their fellow panelists, and UK justices are the least.
dissent rate even if her lifetime liberal voting rate is the equal to the average of all the other panelists. We could think of this dissent rate as the baseline dissent rate as the judge’s voting tendencies are similar to the average of others. Australian justices have comparably high dissent rates. The UK and Canada have similar and much lower base dissent rates.
Second, justices on the US Supreme Court and the UK House of Lords are both relatively unresponsive to being different to the average of the other panelists. As the justice’s voting tendencies become increasingly different from the average of the other panelists (we move to the right in the Figure), the probability of dissent only slowly increases. For Canada and Australia, on the other hand, differences from the panel average are correlated with a much greater increase in the dissent rate. The United States and the UK may have stronger norms—for the United States a strong norm of disagreement may allow justices to overcome at least some of the negative consequences of dissenting (such as reputational effects with her fellow justices), and for the House of Lords a strong norm of agreement may mean that it is difficult to dissent even in the face of a large difference in voting tendencies of the justice and the other panelists. The Supreme Court of Canada and the Australian High Court may have weaker norms, leading to differences in dissent rates being more closely related to differences of the justice from the rest of the panel.
As we discussed, the ideology hypothesis may interact with an important institutional factor: panel size. The larger the panel (and, by extension, the larger the court), the greater the opportunity for ideological differences among judges. This difference is noticeable in India, where panel size has changed over time. When the Supreme Court used to sit in plenary sessions, there was at least one dissent in every freedom of speech case. Once smaller panels were instituted, dissent occurred less frequently.58 Songer, Szmer, and Johnson find that the likelihood of a dissent increases on the Supreme Court of Canada as the panel size increases, with a one standard deviation increase in panel size increasing the odds of a dissent by over 42 percent.59 However, it cannot be assumed that panel size alone is responsible for more pronounced ideological division. Larger panels may be used because a case is particularly divisive.60 Likewise, smaller panels may have fewer dissents not only because there are fewer judges who could potentially disagree, but also because the cases heard by small panels tend to be more routine.61 In any event, these results control for differences in panel size on a particular high court.62 the judge’s lifetime liberal voting rate and the average of the other panelists. These results were all significant at the .01 level except for the UK, which was significant at the .1 level. We also ran the same regressions using the judge’s lifetime liberal voting rate in criminal matters and examining non-criminal dissent rates. The results were similar, except that the Canadian dissent rate did not increase quite as quickly for when the party of the appointer was used, and that the UK did not have a significant change in the dissent rate as the judge became increasing different from the panel average.
- 58. Abhinav Chandrachud, “Speech, Structure and Behaviour on the Supreme Court of India” (2012) 25:2 Columbia Journal of Asian Law 222 at 252. Smyth has remarked that one reason that the Australian High Court produces more separate opinions than either the US or Canadian supreme courts is the panel size: the High Court’s typical panel consists of only five judges. There is therefore less opportunity for large coalitions in Australia; coupled with the seriatim tradition, this makes for fragmented opinions. Smyth, “Some Are More Equal,” supra note 22 at 200.
- 59. Songer et al., “Explaining Dissent,” supra note 22 at 404.
- 60. Songer et al., “Explaining Dissent,” supra note 22 at 404.
- 61. McCormick, “With Respect,” supra note 18 at 107.
- 62. In our dissent regressions on individual judge votes, the panel size was positive and significant at the .01 level for Canada and the UK and insignificant for the other countries. See also Edelman et al., supra note 25, who suggest in passing that there will be more dissent as the panel size increases due to simple probability, and they control for number of judges. See also Alarie et al., “Panel Selections,” supra note 37 for a discussion of panel size.
-  Forrest Maltzman, James F. Spriggs II & Paul J. Wahlbeck, “The Politics of Dissent andConcurrences on the US Supreme Court" (1999) 27:4 American Politics Research 488 at 501.Jeffrey A. Segal & Harold J. Spaeth, The Supreme Court and the Attitudinal Model Revisited(Cambridge: Cambridge University Press, 2002) at 281 (Segal and Spaeth note that assigningthe majority opinion to the most marginal judge (the judge at highest risk of switching over tothe dissent) does not help keep a majority coalition together or increase the size of the majority). See also Forrest Maltzman, James F. Spriggs II & Paul J. Wahlbeck, Crafting Law on theSupreme Court: The Collegial Game (Cambridge, UK: Cambridge University Press, 2000) at 133.They call this the “opinion distance hypothesis."
-  Walker, Epstein & Dixon, supra note 8 at 375. See also Sunstein, “Unanimity," supra note 3.
-  Haynie, “Leadership,” supra note 16 at 1166.
-  Edelman et al., supra note 25.
-  At the state supreme court level, Leonard and Ross find that ideological diversity does nothave a significant effect on the likelihood of unanimity, although elected judges show higherrates of dissent: Leonard & Ross, supra note 39 at 14, 18.
-  Songer et al., “Explaining Dissent,” supra note 22 at 401. As discussed further in the subsequent sections, they also find that the presence of interveners, certain case issues (constitutional, civil liberties, and statutory interpretation), panel size, reversal, and complexity, alongwith being heard during Lamer’s tenure as chief justice (relative to Laskin) were positively andstatistically significantly related to the dissent rate. Neither the fact that the issue related tocriminal law, or workload, or having Dickson or McLachlin as chief justice were statisticallysignificant. See also Donald R. Songer, The Transformation of the Supreme Court of Canada: AnEmpirical Examination (Toronto: University of Toronto Press, 2008) at 86.
-  Russell Smyth, “Explaining Voting Patterns on the Latham High Court” (2002) 26:1Melbourne University Law Review 88 at 106. See also Schubert’s early study of agreement on theHigh Court concludes that differences in opinion agreement cannot be accounted for based on
-  We follow Epstein, Landes & Posner, Behavior, supra note 3, who use fraction dissenting toexplore the determinants of dissenting.
-  See Appendix 3 for the regressions underlying this Figure. For this Figure, we used the estimates from regressions fraction dissenting on TotalCases, time, and idsd. We held TotalCases atits mean for each country and time at the median. The marginal values for idsd was significantat the .05 level for each country.
-  See Appendix 3 for a discussion of the variables used and the results of the relatedregressions.
-  We also examined whether a judge’s dissent rate appeared correlated with both whetherthe judge was on a homogeneous panel (that is, all judges appointed by appointers from thesame party), and whether the judge was a minority on the panel in terms of ideology (such as,for example, the sole liberal appointee on a panel where all the other judges were appointed bya conservative appointer). Overall, these results showed no significant difference in the probability of the judge dissenting.
-  These results were found using data on individual judge votes for each country. The dependent variable in each case was whether the judge dissented or was with the majority. The independent variables included the party appointing the judge, the lower court direction (liberal orconservative), the area of law, the panel size, and a time trend, as well as the difference between