June 28, 2022

Factors Influencing the Likelihood of Winning an Appeal at the Supreme Court of Canada

Earlier this year, we launched our Supreme Court of Canada Decisions Project. Our dataset contains information about every Supreme Court of Canada decision going back to the mid-1950s.

In addition to our own internal uses, we are releasing a series of blog posts showing some of the insights we have gleaned from analyzing our data. Our first blog post provided some summary data about the evolution of the Supreme Court over time. In this blog post, we look more particularly at a question of interest to many: what factors make you more likely to win your appeal at the Supreme Court of Canada?

For this exercise, we will not be looking at the data all the way back to the 1950s. Instead, we focus on a more recent time period: the last 10 years. Given that our dataset goes to the end of 2021, this means that we include all appeals decided between January 1, 2012 and December 31, 2021.

In addition to the temporal limitation, we imposed two further limitations on the data for our analysis. First, we included only appeals – motion decisions and Federal references were excluded. Second, we excluded a handful of cases where there was an unusual disposition, such that there was not a clear win for either the appellant or respondent.

There were 659 decisions rendered during this period that met those criteria. Of those decisions, approximately 45% were won by the appellants, and 55% were won by the respondents. (In this context, a case is labelled as being won by the appellants if the appeal was allowed in whole or in part.)

Using the data contained in our Supreme Court database, we then looked at whether the likelihood of an appellant being successful varied based on a number of factors. Among other things, we looked at the likelihood of success based on:

  • features of the lower court decision (e.g. whether the appeal was allowed at the lower court and whether there was a dissent);
  • features of the parties (e.g. whether the appellant was the federal government, an individual, or a corporation);
  • the issues raised by the case;
  • the court being appealed from;
  • whether the case came to the court by right or by leave.

We did not necessarily expect to find significant patterns among these variables. To the contrary, while our prior work on the Supreme Court of Canada Leave Project had shown us that factors like these can play an important role in the Supreme Court granting leave to a case, we were more sceptical that these types of factors would influence the probability of an appeal being allowed. The determination of success on appeals to the Supreme Court of Canada, we thought, would be based more on aspects of legal reasoning, and the merits of the decision below, that our dataset simply could not pick up.

As it turned out, we were wrong. Our analysis identified a number of factors that were statistically significantly corelated with the likelihood of an appeal being allowed.

We started with a univariate analysis (i.e. looking at the probability of an appellant winning their appeal, based on the presence or absence of individual factors, without consideration of other confounding factors). The dot chart below shows the probability across all cases of the appellant winning, as well as the probability of the appellant winning in cases which contained one particular feature.

Because univariate results can be misleading, we also ran a series of multivariate logistic regressions to assess whether the factors we identified still had meaningful effects and were statistically significant after accounting for other factors. Based on both analyses, here is what we found.

The Relevant Factors

First, appellants in cases that were heard after the Court granted leave are more likely to win their appeals than appellants in cases where the Supreme Court heard the case as of right. The Supreme Court allowed approximately 47% of cases in which leave was granted, while it allowed just 40% of cases in which the appeal was as a right. (The magnitude of this factor is relatively small compared to some of the other ones, but it does remain statistically significant, even after controlling for other factors. Indeed, all of the factors identified in this section remained statistically significant, even after controlling for other factors.)

This outcome is not surprising. The test for getting leave to the Supreme Court is whether the case raises an issue of public importance. However, it is reasonable to expect that one factor that goes into the Supreme Court’s analysis is with regard to whether the decision below was right or wrong. Intuitively, Supreme Court judges might be more willing to grant leave where they see an error to correct, while they play no such screening role in cases that come to the Court as of right. And indeed, this analysis suggests an error correction may play a role in the leave decision.

Second, certain features of the decision of the court below substantially impacted the likelihood of an appeal being successful. Both a dissent at the Court of Appeal and the Court of Appeal allowing the appeal were both statistically significantly associated with a higher likelihood of the appellant being successful. Appellants were successful at the Supreme Court of Canada in 51% of cases with dissents at the Court of Appeal (compared to 42% in cases with no dissent), and 54% of cases in which the Court of Appeal allowed the appeal (compared to 36% in cases where the appeal was dismissed).

There are several reasons why these features of a Court of Appeal decision may be corelated with the likelihood of a successful appeal by an appellant. The presence of both an appeal below being allowed, as well as a dissent, indicates there was some disagreement among courts below. That disagreement may provide a road map to both appellants and the court as to how to overturn the Court of Appeal’s decision. Alternatively, part of the explanation may relate back to the leave stage. If, as we speculated above, error correction does play a role in deciding to grant leave to cases, it may be that the court is more likely to grant leave on an error correction basis when it has a clear path to identify that error from either a decision at first instance or a dissent that articulates the errors in the reasoning of the majority of the Court of Appeal.

Third, there appears to be some evidence in our dataset that the presence of a concurring decision at the Court of Appeal substantially reduces the likelihood of an appeal succeeding at the Supreme Court of Canada. Over the 10-year period we looked at, appellants were successful at the Supreme Court of Canada in just 29% of cases where there was a concurrence at the Court of Appeal.

The presence of why a concurrence might lower the chance of winning at the Supreme Court is fairly straightforward. It might mean that while several judges arrived at the same result at the Court of Appeal, they arrived there by different paths. That in turn suggests that there may be multiple paths to victory for the respondent at the Supreme Court of Canada, which intuitively might increase their chances of victory.

We note caution around this conclusion. The number of Court of Appeal decisions with concurrences is relatively small, so the sample size of cases analyzed there is a modest one. Moreover, while our analysis for the 10-year period of 2012 through 2021 showed a statistically significantly increased likelihood of the appellant being unsuccessful in the presence of a concurrence, that feature disappeared when we took a broader look at a 20-year period arising back to 2002. By contrast, the other variables we describe here remain statistically significant. Consequently, we note the possibility of an interesting relationship here, although we do not wish to be taken as definitive in that relationship.

Fourth, we found that appellants had a substantially increased chance of being successful in private law appeals than other types of appeals. Appellants were successful in 57% of private law appeals to the Supreme Court of Canada, while appellants in other cases were successful in just 43% of appeals.

This factor again may again be explained be a selection effect at the leave stage. Criminal and public law cases may more obviously raise issues of national importance, so the Supreme Court may be inclined to hear them, even if they are inclined to dismiss the appeal. By contrast, in private law cases, the fact of an error below may be one factor that motivates the court to grant leave in the first case, which in turn would corelate to a higher likelihood of the appellant being successful.

Fifth, we found that the appellant being the federal government increased the likelihood of an appeal being allowed. Where the federal government was the appellant, it won in approximately 63% of cases. However, it is important to note that in our dataset, the Crown in prosecutions under the Criminal Code, is coded as the federal government. Consequently, while this suggests that the federal government is more successful when it appeals, what this actually indicates is that Crown appeals tend to be more successful on average.

Sixth, while this is not a factor that a party can know in advance of a hearing, we did find that appeals decided from the bench were less likely to be won by appellants. Where a case was decided from the bench, an appellant had only a 38% chance of succeeding. By contrast, where the Supreme Court reserved its decision, appellants won almost 48% of the time. This means the very fact of the court deciding to reserve its decision in the case increases the likelihood of the appellants succeeding.

The Irrelevant Factors

While the most interesting factors are the ones that are relevant, we also flag a variety of factors that we did not find had any statistically significant relation to the likelihood of the appeal being allowed.

First, other than at the high level of private law versus public law versus criminal law, we did not find that the Supreme Court of Canada was more likely to allow an appeal that raised particularly more narrowly defined areas of law than others.

Second, we did not find statistically significant differences in the rate of appeals being allowed from appellate courts versus others, after controlling for other factors. Using a purely univariate analysis, and before accounting for statistical significance, it did appear that the rates of appeal being allowed varied by court of origin. However, this largely appeared to be a function of small sample sizes for appeals that came from certain Courts of Appeal. Most courts with relatively large numbers of appeals to the Supreme Court of Canada (i.e. British Columbia, Alberta, Ontario, and Quebec) were relatively close to the average rate of appeals being allowed. Once we incorporated these factors into a multivariate multiple, we found no statistically significant variations by Court.

What Do We Do With This Information?

All of these factors are interesting as an intellectual exercise, but what is the practical relevance of this analysis to lawyers?

First, these insights are extremely valuable in simply understanding the Supreme Court as an institution. That has inherent value in itself.

More practically, a potential appellant’s decision as to whether to seek leave to appeal to the Supreme Court should be influenced not just by their chances of winning leave, but by their chances of succeeding in the appeal. The chances of winning an appeal to the Supreme Court of Canada may also implicate the probability of settlement of a dispute. Consequently, accurate estimates of the likelihood of winning an appeal at the Supreme Court of Canada can help litigants make better decisions.

This is especially true because of the difficulty in predicting outcomes at the Supreme Court of Canada using conventional legal analysis. Cases in which the Supreme Court has granted leave are likely to be the cases that are the most difficult to predict the outcome of using conventional legal reasoning. Lawyers are used to considering cases by evaluating legal decisions by whether the applicable legal principles can import with binding precedent. However, at the Supreme Court level, there is no higher court to provide a binding precedent, and the Court is more free to either tweak or depart from its existing precedent.

Because of these factors, it can be more difficult to predict the outcome of Supreme Court appeals using conventional legal analysis than it is to predict the outcome of appeals to lower courts. In turn, these models can improve on the basic sense that one has a 50/50 shot at the Supreme Court by giving a better estimate based on case-specific factors.