The Federal Government's Lawyer: The Attorney General of Canada

According to the Department of Justice Act, first passed in 1868, the Attorney General of Canada (hereafter AG Canada) “shall have the regulation and conduct of all litigation for or against the Crown or any department, in respect of any subject within the authority or jurisdiction of Canada” (Department of Justice Act, R.S., 1985, c. J-2, s.5). With few exceptions,Footnote ⁴ the AG Canada does indeed exercise this monopoly over litigation, as well as providing legal advice to the government, while also sitting simultaneously in Cabinet as the Minister of Justice with responsibility over policy matters relating to the justice system. As Reference HennigarHennigar (2002) notes, however, the AG Canada does not personally conduct the government's litigation. This task falls primarily to legal-area specialists in several Regional Offices across Canada and, less frequently, Justice Department lawyers located in various line departments and in the Department's headquarters in Ottawa.Footnote ⁵ This is true even of appeals to the SCC, as the original counsel from lower court cases are replaced less than 25 percent of the time (Reference HennigarHennigar 2002:96). Thus, there is no Canadian counterpart to the Office of the Solicitor General in the United States, that small group of elite lawyers responsible for representing the national government before the Supreme Court.

The Appeal Process

When the Canadian government loses in either the federal or provincial penultimate courts of appeal,Footnote ⁶ it has fewer appellate options than its American counterpart. Reference ZornZorn's (2002:146–7) study of the U.S. government's appeals from the federal courts notes that government lawyers have four options, if the lower court decided the case with the typical three-judge panel: (1) concede, and forgo any further action in the case; (2) request a rehearing by the same three-judge panel, which he notes is rare; (3) request a rehearing en banc (that is, by all members of that Circuit Court of Appeals); or (4) appeal to the U.S. Supreme Court, by filing a petition for a writ of certiorari. In contrast, the AG Canada has really only two options following a loss in the penultimate courts of appeal: concede, or appeal to the SCC. Despite the fact that the penultimate appeal courts sit in three- (or very rarely, five-) judge panels drawn from considerably more members (24 in Ontario, for example), there is no parallel provision authorizing a request to rehear a case en banc, nor by the original panel.

The SCC has enjoyed virtually full docket control since 1975, with almost all appeals to the Court requiring its permission, or “leave to appeal” in the Canadian parlance. With some exceptions,Footnote ⁷ the appeal process is governed by Section 40(1) of the Supreme Court Act, 1985, which Reference Flemming and KrutzFlemming and Krutz characterize as “elastically worded and vaguely defined” to maximize judicial discretion (2002:233). Similar to Rule 10 governing appeals to the U.S. Supreme Court, Section 40(1) instructs the Court to grant leave when “the Supreme Court is of the opinion that any question involved therein is, by reason of its public importance or the importance of any issue of law or any issue of mixed law and fact involved in such question, one that ought to be decided by the Supreme Court or is, for any other reason, of such a nature or significance as to warrant decision by it.”

Although the government's legal representation is somewhat decentralized geographically and across policy areas, its litigation strategy regarding appeals to the SCC level is highly centralized. These strategic issues include the decision whether to appeal and the authoring of facta, or written legal arguments presented to the Court. As a senior Justice Department official put it,

Formal Ministry approval for SCC appeals includes review and recommendation by the National Litigation Committee (NLC), with the final decision by the AG Canada. The NLC is composed of several senior Justice Department lawyers, who are responsible for “reviewing all recommendations to appeal or not to appeal significant cases” (Department of Justice Canada 2000: Sec. 46.3.6). Every case in this study qualifies as “significant” under the NLC's guidelines (Department of Justice Canada 2000: footnote 4, at Sec. 46.4), as they each involve a constitutional (Charter) question that may be appealed to the SCC. As such, the NLC would have made a recommendation in every case studied here. The NLC recommends appeal only where “the public interest requires an appeal” (Department of Justice 2000: Sec. 22.3; emphasis in original), a flexible standard that includes the following considerations:

• • Is the issue raised by the case of widespread importance, or is its impact confined largely to the immediate case? In Reference ZornZorn's terms (2002), this corresponds to a case's salience.

• • Have the courts differed in interpreting the issue raised?

• • Could the decision impair the enforcement or administration of a significant government policy initiative of the law if left unchallenged?

• • Will the resources required to prepare and present the appeal significantly outweigh the value of pursuing the case further (a cost factor)?

• • Only where the arguments for and against appealing are evenly matched, public expressions of concern may tip the scales in favor of an appeal.

The final stage in the appeal decision process rests with the AG Canada/Minister of Justice and his or her deputy. Ministerial approval of appeals to the Supreme Court is no mere formality. According to a long-standing member of the NLC, “The [National Litigation] Committee always treats its work as a recommendation to the Deputy and to the Minister … . There is further review, and the committee is very cognizant that the Minister has the last word” (Personal interview, Robert Frater, Senior Counsel, Department of Canada, 17 April 2002, Ottawa). Moreover, at this level, the Justice Minister's decision may be influenced by the political interests of his or her Cabinet colleagues, including the Prime Minister. Another NLC member recalled instances where the Committee concluded there was an issue of public importance, but left the decision entirely with the Minister due to the presence of “other, more political concerns where we have no expertise.” (Personal interview, Graham Garton, 31 July 2001, Ottawa). For example, during the recent litigation involving same-sex marriage, sources inside the Liberal government revealed that the conduct of the lower court cases was debated at the highest levels of the political executive, by the Cabinet and then-Prime Minister Jean Chrétien, and included calculations about the issue's impact in an anticipated election (Reference MacCharlesMacCharles 2004:H1, H4). In short, at this level, senior political figures may influence the decision to appeal.

Explaining Government Appeal Decisionmaking: Hypotheses and Operationalization

As Reference ZornZorn (2002:149) notes, an analysis of appeal decisionmaking must begin by asking what the government's goals are when appealing. Most fundamentally, we can assume that—like any appellant—the government seeks to minimize or reverse the loss it has suffered in the lower courts. All losses in litigation entail some cost to the losing party, and when the government loses in court, its costs can be sorted into two categories: financial and policy costs. Like a private litigant, the government may face financial costs if found liable in civil suits, as well as the cost of the resources required to prepare and present the case. In addition, however, the state may bear fiscal costs associated with policy change; for example, a decision extending state-funded “maternity” leave benefits to fathers (R. v. Schachter 1992) would, left unaltered, significantly increase the government's expenditures in that program. Actual financial costs defy measurement, but as Reference ZornZorn (2002:153–4) reasons, following Reference GalanterGalanter (1974), civil litigation is typically more costly than criminal prosecution. As well, a centralized appeal decisionmaking process, as exists in the United States and Canada, should be less concerned with criminal case losses “because the costs of a court of appeals reversal in a criminal case is [sic] borne by someone other than the individual(s) responsible for making the final appeal decision” (Reference ZornZorn 2002:154). Accordingly, I hypothesize that the AG Canada will be more likely to appeal in civil cases than in criminal cases. As such, I created a variable where criminal cases (cases when the federal government initiated prosecution of an offense) were scored as 0 and civil cases scored as 1, with the expectation that a positive coefficient would emerge on this civil litigation variable.

Policy costs include the loss of a particular statute and may also entail a diminished capacity to regulate that policy area more generally. A related policy cost is the opportunity cost and resources (in time, personnel, and political capital) associated with attempting to draft replacement legislation. Zorn's measure for policy costs—whether the court has invalidated a federal law or statute—is logical, but incomplete: a more accurate measure would include all forms of constitutional remedy: invalidation (nullification), constitutional exemption,Footnote ⁸ severance,Footnote ⁹ and “reading in” (judicial extension). These can be sorted into two categories of judicial remedies, based on the degree of judicial intrusiveness. As one senior Justice Department official stated,

To capture this distinction I created two variables: “less-intrusive remedy” and, for more intrusive remedies, “judicial amendment.” As severance and reading in both entail outright judicial rewriting of the law, penultimate appeal court cases in which either occurred were coded 1 on the judicial amendment variable, with all other cases coded 0. However, constitutional exemption leaves the law essentially intact, and invalidation leaves rewriting the legislation (or not) to the government. I scored the case 1 on the less-intrusive remedy variable if either of these remedies were employed by the penultimate appeal court, and 0 if not. While both variables are expected to have positive coefficients, I hypothesize that judicial amendments are more likely to provoke a government appeal than less-intrusive remedies.

As a strategic (that is, institutionally constrained) litigant, the AG Canada must be a selective appellant. As noted earlier, Reference ZornZorn (2002) posits that, under similar conditions, the U.S. solicitor general's choices will be shaped by his or her assessment of what cases the U.S. Supreme Court will consent to hear (a “reviewability” factor) and which can be won on the merits (a “winnability” factor). Reference ZornZorn's (2002:154) primary measure of reviewability is litigant claims of intercircuit conflict, reflecting the importance of this factor among the U.S. Court's criteria for granting certiorari. Flemming's recent study of SCC leave to appeal decisions found that “a claim of conflicting lower court decisions, as the American literature amply suggests, is more likely to be heeded by Canada's Court than an argument lacking this trump card” (2004:68). Mirroring the SCC's behavior, the Department of Justice Canada's guidelines (2000: Sec. 23.2.1) explicitly cite jurisprudential conflict between appeal courts as a matter of “national importance” warranting appeal. The Justice Department also has its own reasons to appeal because of intercourt conflict. As a “repeat player” with a national focus, the federal government seeks legal coherence and consistency across regional jurisdictions. As one senior department official explained, “We consider it generally not a very good state of affairs if [federal government] lawyers in one province are precluded from taking a position that they might be able to take in another province” (Personal interview, Robert Frater, 17 April 2002, Ottawa). An “intercourt conflict” variable was created to capture this factor, with cases coded as 1 if the penultimate appeal court cited conflicting lower court precedents, or where a precedent from a court of equal or higher rank was cited by that court but not followed; cases without such citations were coded as 0. As the government should be more likely to appeal when the lower appellate court cites intercourt conflict, I expect this variable to exhibit a positive coefficient.

A selective appellant is also likely to focus its resources and limited opportunities on cases it considers more important, or “salient.” Zorn only links case salience to reviewability (on the argument that the Court is more likely to hear more important cases), but it is appropriate to conceptualize salience as a separate factor influencing government litigants. Zorn identifies three indicators that a case is particularly important, and therefore more likely to be appealed: a constitutional issue is raised, the lower court's decision is published, or an amicus curiae (third-party “intervener,” in the Canadian parlance) appears in the lower court (2002:154–5). Following this logic, only the presence of an intervener is applicable to my study, as every case involves a constitutional (Charter) question, and, for the pragmatic reasons cited in the next section, only published decisions are examined. Although interventions by interest groups and governments are common in the SCC's Charter cases (Reference BrodieBrodie 2002), they are relatively rare in Canada's lower courts. Their presence is therefore a good indicator that the case involves an issue important enough to provoke legal mobilization by political actors not directly involved in that case. Interveners also indicate that the case has attracted “public attention,” which the NLC guidelines cite as a factor favoring appeal. I coded the intervener variable 1 if an intervener was present in the lower appeal court case and 0 if not, with the expectation that this variable will be positively related to appealing.

A case may also be more salient because its impact is not—to echo the Justice Department's guidelines—“confined largely to the immediate case”; that is, the lower court decision has a broader scope. An important dimension of this is the ruling's impact on the elected government's power relationships to other state institutions, particularly the judiciary. As Reference SalokarSalokar phrases it in her study of the U.S. solicitor general, “Has the executive branch, as a result of the lower-court decision, been weakened vis-à-vis the power of other branches, and to what degree has this occurred? If the holding results in a significant decrease in executive power … the case will receive serious consideration for submission [to appeal]” (1992:11). Unlike Zorn, I measure the scope of a lower court decision directly, by focusing on novel legal rule development. When, through the interpretation of constitutional provisions, a lower appeal court crafts a new legal rule that will erode the government's authority in other policy fields, or vis-à-vis other state actors or individuals, the government should be more likely to appeal the decision. I scored the “novel interpretation” variable dichotomously, 1 for cases with such rulings—which include the application of the Charter to new issues and jurisdictions, the creation of new judicial remedial powers, the augmentation of judicial discretion over the application of the Charter, and expanding the scope of justiciable rights—and 0 otherwise. A legal interpretation was considered novel if no Canadian precedent was cited for the rule.

Preliminary analysis revealed a conditional relationship between the two measures of salience, in that novel interpretation was only statistically significant in the presence of an intervener. To capture this relationship, I followed Reference WrightWright (1976) and created two conditional variables to replace the novel interpretation variable in multivariate analysis, while retaining the dummy variable for interveners. The first (“Novel Interpretation & Intervener” in Table 1) was coded 1 if both novel intervention and an intervener were present in the lower court case, 0 otherwise, and is expected to have a positive coefficient. The second (“Novel Interpretation & No Intervener”) was coded 1 if a novel interpretation was present but in the absence of an intervener, 0 otherwise. While it should also have a positive coefficient, it is not expected to obtain statistical significance.

Table 1. Summary Statistics, Determinants of Appeal Decision by AG Canada

^† When binary, proportion of cases variable=1.

The final strategic factor considered here is the prospect of winning on appeal. Reference ZornZorn (2002:152) contends that governments will not waste resources or risk their credibility with the U.S. Supreme Court by litigating “lost causes.” Winning is also important for jurisprudential reasons, namely, that losing before the country's highest court of appeal may establish an unfavorable legal rule with the widest possible application. Waltenburg and Swinford's survey found that “the states are well aware that litigation entails risk. Namely, the state interest might lose, and even worse, this disadvantaged position is then bolstered by the permanency accompanying a Supreme Court ruling. As one state respondent explained, ‘We will not put good precedent at risk’” (1999a:254). The risks of appealing are even greater for the Canadian federal government, in that its losses are limited when they occur in provincial courts of appeal, while a loss before the SCC has consequences for the application of federal law at the national level. This is because rulings by PCAs (as opposed to the Federal Court of Appeal [FCA]) apply only in the jurisdiction of that court (for example, Ontario for the Ontario Court of Appeal), but the SCC's decisions apply in all jurisdictions.Footnote ¹⁰ In view of this risk, the federal government have an extra incentive to avoid losing in the SCC, although this factor should be more prevalent in cases originating in PCAs than in the FCA.

Zorn operationalizes winnability in three ways: whether there were reversals among the lower courts as the case was appealed, whether there was dissent on the highest court of appeal before the U.S. Supreme Court, and whether the U.S. Supreme Court's ideological orientation favored rights claims. All of these measures are sound, and I adopt them here. Reversals suggest that the government was successful at some level, and a lower court dissent indicates that the government persuaded at least one senior appellate court judge to adopt its legal argument. Furthermore, Reference McCormickMcCormick (1994:88–9) found that the SCC overturns the penultimate courts more often when the latter reverse a lower court or dissent, than when the penultimate courts affirm or are unanimous. For the lower court reversal and dissent variables, the presence of the factor (coded 1, 0 if not present) should encourage an appeal and thus have a positive coefficient. To operationalize the ideological orientation of the SCC, I calculated the rate (scored from 0 to 1.0) of the Court's Charter conservatism—or, more accurately, judicial restraint—during the two years before each decision to appeal; the rate reflects the proportion of rulings that denied rights claims. This provided a “moving” measure of the SCC's recent treatment of Charter claims at the time the AG Canada had to decide whether to appeal. Moreover, by focusing on the Court's tendencies as a whole, this approach sidesteps the measurement problems associated with varying panel composition (which ranges from five to nine justices) and the existence of ideological cleavages within the Court (Reference HeardHeard 1991; Reference McCormickMcCormick 1999; Reference OstbergOstberg et al. 2002; Reference Songer and JohnsonSonger & Johnson 2002; Reference Tate and SittiwongTate & Sittiwong 1989; Reference Wetstein and OstbergWetstein & Ostberg 1999). As I hypothesize that the AG Canada will be more likely to appeal to the SCC when that Court's recent track record indicates less support for rights claimants (bearing in mind that, by definition, the government's “opponent” in the case is a rights claimant), this SCC conservatism variable should have a positive coefficient.

Finally, I included a variable to test the rival hypothesis that the government's appeal decisions are influenced by which political party controls the AG Canada's office. Only two parties, the Liberals and the Progressive Conservatives (Tories), have formed the government since 1982. The adoption of the Charter that year was the culmination of Liberal Prime Minister Pierre Trudeau's constitutional reform agenda, which he pursued for more than a decade. The federal Justice Minister at the time, and Ottawa's chief negotiator with the provinces, was Jean Chrétien, who served as Prime Minister from 1993 to 2003. Under the Tory governments of Brian Mulroney from 1984 to 1993 (and briefly Kim Campbell in 1993), Ottawa pursued constitutional reforms in the Meech Lake (1987–1990) and Charlottetown (1992) Accords, which, while aimed at securing the Quebec provincial government's adoption of the 1982 Constitution, were seen by many as weakening the Charter's rights (Reference CairnsCairns 1992; Reference RussellRussell 2004).Footnote ¹¹ As such, I hypothesize that the federal government will be more likely to challenge its losses in Charter cases when the Tories are in power. In such cases I coded the Conservative government variable 1, 0 if a Liberal government, and expect a positive coefficient. Given the Tories' long reputation as a stronger “law and order” party, I further hypothesize that they will be more likely than the Liberals to appeal their losses in criminal cases.

The variables are summarized in Table 1. Because the dependent variable, appeal (=1) and not appeal (=0), is a dummy variable, all estimation is by maximum likelihood (MLE) using binary logistic regression. Notably, the categories listed below do not exhaust possible motivations, nor are they watertight compartments. For example, an invalidation or judicial amendment would also signify a case's greater salience, and dissent on the lower court improves the odds that the SCC will hear the case (reviewability). Rather, they should be considered as “heuristic categories” that help organize the “diverse set of elements which enter into the appeal calculus” (Reference ZornZorn 1997:75).

Data

The following case selection criteria were employed to construct the database of cases requiring a decision whether to appeal:Footnote ¹²

1. 1 the case was a reported decision of a PCA or the FCA (see footnote 6 above),

2. 2 involved a claim under the Canadian Charter of Rights and Freedoms

3. 3 decided between 1982 and 2000 inclusive,

4. 4 the federal government was a party (appellant or respondent) to the case, and

5. 5 the federal government lost the dispute (i.e., its appeal was dismissed, or the appeal was allowed when the federal government was the respondent) in the penultimate court of appeal.Footnote ¹³

Reported decisions were used because the detailed case information required for the study is typically unavailable for unreported decisions. The Canadian Abridgment, a comprehensive electronic list of all Court of Appeal and SCC decisions, was used in conjunction with major national law reports and several provincial reports to compile the list of reported cases. The study's start date of 1982 marks the adoption of the Charter of Rights and Freedoms. Notably, this time frame means that the cases used for the study are not a sample but the actual universe of Charter cases (to 2000) involving the federal government that meet the remaining criteria. As such, there is no selection bias in the cases studied.

The first four case selection criteria generated 593 cases, of which the federal government lost only 160—a remarkable 73 percent overall success rate on case disposition in the penultimate courts of appeal.Footnote ¹⁴ The government was only slightly less successful in civil litigation (168 wins of 237 cases, or 70.9 percent) than in criminal prosecutions (267 wins of 356 cases, or 75 percent). The AG Canada's high “net advantage” (Reference WheelerWheeler et al. 1987) of +46.5 percent in lower appeal court Charter cases is comparable to the U.S. government's +45.1 percent in the U.S. courts of appeal (Reference Songer and SheehanSonger & Sheehan 1992). Thanks to its remarkable success rate in Charter cases, the Canadian government has to make relatively few appeal decisions each year (the data reveal no clear trend with respect to time). Nonetheless, it does not appeal such losses routinely, appealing only 48 times (30 percent) between 1982 and 2000, 43 by leave and five by right. Surprisingly, the existence of a right to appeal did not increase the likelihood of appealing. Of the 17 cases where the right existed, only five were appealed (29.4 percent), compared to 43 of 143 potential applications for leave (30.1 percent). Notably, the federal government enjoys a high level of success at securing leave to appeal from the SCC: of 43 applications, all but seven were allowed (83.7 percent success rate). Although the government's behavior implies a certain degree of selectivity, the federal government's appeal rate was considerably higher than that of all classes of appellants from penultimate appeal courts, which has been around 12 percent in recent years.

Regarding the lower court origin of the case, the government was slightly more likely to appeal a decision of the FCA (19/55=34.5 percent) than of a PCA (28/104=26.9 percent). The Justice Department's appeal decisions also varied according to the Charter right claimed. Among the cases lost by the federal government at the penultimate court level, legal (criminal due process) rights claims (Sections 7–14 claims) were by far the most common (n=129) but were appealed at one of the lowest rates (27.1 percent). By contrast, Ottawa appealed almost 60 percent of its losses in Section 15 equality rights cases (10 of 17), and 42 percent (five of 12) of those involving the “fundamental” freedoms of expression, religion, assembly, and association in Section 2, the second- and third-largest categories, respectively. This is not to say that the government's appeal decisions were driven by the Charter right in question; rather, a more compelling interpretation is that equality rights cases tend to contain some of the factors discussed above that encourage appeals, while legal rights cases (arising in criminal prosecutions) do not.Footnote ¹⁵

Results and Analysis

The results from logistic regression are summarized in Table 2. Overall, the model represented a good statistical fit to the data, producing a 23 percent improvement in prediction over the null model; the p value of 0.532 in the Hosmer-Lemeshow test implies that the model's estimates fit the data at an acceptable level. The Nagelkerke pseudo-R² of 0.31 indicates that the overall model hypothesized here was fairly strongly associated with the decision to appeal. However, when employing a 0.05 criterion of statistical significance, only three explanatory factors—judicial amendment, the joint presence of a novel constitutional interpretation and intervener, and dissent on the lower appeal court—had significant effects. I discuss each category of factors in turn below.

Table 2. Logit Model of Decision to Appeal by AG Canada

Note: The column entries are unstandardized regression coefficients, with standard errors shown in parentheses. Estimation is by maximum likelihood using binary logistic regression.

^* p<0.05,

^** p<0.01^a,

^† Predicted effect on probability that government appeals (y=1) of changing each explanatory factor from its minimum to maximum (i.e., absent to present for dichotomous variables) when all other variables are held at their means (modes for dummy variables).

^a Where the study uses the population of cases rather than sample data, p values are technically irrelevant. Nevertheless, they provide a useful and widely understood metric for evaluating results.

The hypothesis regarding the influence of financial costs is not supported. Although losses in civil cases were, as hypothesized, more likely to be appealed than those in criminal cases, the factor did not emerge as statistically significant in the multivariate model. The hypotheses regarding policy costs fared somewhat better, particularly that predicting the greater influence of judicial amendment (“reading-in” and severance) compared to the less-intrusive remedies of invalidation and exemption. Nine of 13 cases involving judicial amendment by the lower appeal court (69.2 percent) were appealed to the SCC, compared to 13 of 33 invalidations (39.4 percent). Furthermore, while judicial amendment was among the strongest positive predictors of a government appeal (the odds ratio indicates that when holding all other variables constant, an appeal is 5.4 times more likely in the presence of judicial amendment than when absent), there was no statistically significant relationship between less-intrusive remedies (or even invalidation by itself) and the dependent variable. The evidence refutes the claim of some interest groups (Reference Shilton, Monahan and FinkelsteinShilton 1993)—and more interestingly, some government lawyers (Reference Mitchell, Monahan and FinkelsteinMitchell 1993; Reference Gallagher, Monahan and FinkelsteinGallagher 1993)—that the government typically defends impugned legislation, and suggests a more nuanced governmental response to judicial activism. In light of the comments of government officials cited earlier, the government's greater dissatisfaction with judicial amendment than with invalidation is probably due to the degree of freedom the remedy gives the government to respond legislatively. Invalidation offers the most freedom in this respect, while either form of amendment offers the least, and as such, appealing a judicial amendment may offer an easier way to “undo the damage” to governmental policies than a formal legislative amendment. This finding alone suggests that government appeals are, at least to some extent, the product of calculated decisionmaking by central officials concerned with the levels of judicial and governmental power.

Case salience factors—novel legal interpretations (scope) and interveners (third-party legal mobilization)—exert the strongest influence on the government's decision to appeal, but this influence is curiously conditional on their mutual presence. In other words, novel interpretations by the lower courts only provoked appeals when an intervener was also present. The robust results for the conditional variable “Novel Interpretation & Intervener” in Table 2 reflect this finding. Of the cases containing both of these features, only one was not appealed. By contrast, “Novel Interpretation & No Intervener” did not achieve statistical significance. The explanation for the findings regarding case salience is less obvious. One possible interpretation is that the presence of an intervener signals to government lawyers that a case involving a novel interpretation is of particular importance and should therefore be appealed. A second is that the government is inclined to appeal in cases involving particularly important novel interpretations, and that these are the types of cases that also attract interveners. In other words, the same factor that encourages the government to appeal—case importance, tied to novel interpretation—also draws interveners, but intervention is not causally related to appealing. It is not possible, using statistical methods, to determine which interpretation is more accurate, but one conclusion is consistent with both—intervention and novel interpretation are cumulative indicators of case importance, as an appeal is much more likely in the presence of both factors than of either factor by itself. Notably, this conditional relationship appears to be unique, as tests revealed no significant interactions between interveners and judicial amendment and lower court dissent.Footnote ¹⁶

The divergent findings for novel interpretation and judicial amendment are, however, somewhat puzzling. Both measures tap a similar motivation for government appeals: to defend policymaking authority from judicial intrusion. This being the case, it is surprising that novel interpretations are not a significant influence on their own, while judicial amendments are. The explanation may lie with the profile or “visibility” of each with government decision makers. Judicial amendments obtain a high profile, as legislation is necessarily altered, thereby drawing the attention of political officials responsible for the law or policy, as well as that of bureaucrats who administer the law and those who are directly affected by it. These vested interests are almost certain to bring the matter to the attention of senior Justice Department officials, either through the Minister of Justice, government lawyers located in other ministries, or issue-specific units in Justice Headquarters, or when consulted as clientele by the NLC. By contrast, novel interpretations may not decide the ultimate outcome of the case and, as a consequence, they can be difficult to track for vested interests. Thus perhaps only the most important novel interpretations achieve the high profile necessary to provoke an appeal. Notably, this explanation is consistent with the argument, outlined above, that there is no causal relationship between intervention and novel interpretation, but that particularly important cases—that is, those in which major novel interpretations are likely—both attract interveners and spur appeals.

There were several reasons to expect that Justice Department decision makers would be concerned with the reviewability factor of gaining access to the SCC, including the Court's docket constraints and a desire to resolve legal indeterminacy associated with conflicting lower court rulings. It is surprising, then, that this factor enjoyed only weak statistical support, as it had influence in the expected direction but fell just short of reaching statistical significance (p=0.11). However, an interesting distinction emerges when one controls for whether the case was heard in the FCA or a PCA. Citation of intercourt conflict did significantly increase the likelihood of appeal from the provincial courts (b=1.535, p=0.016, odds ratio=4.64) but was not a remotely significant predictor of appeals from the FCA (b=−1.045, p=0.415). This makes some intuitive sense, as intercourt jurisprudential conflicts are more likely among the 10 provincial (and, during this time period, two territorial) courts. Among the Charter cases lost by the federal government in this study, 19.3 percent of those in PCAs cited intercourt conflict, compared to only 11.8 percent of those in the FCA. More important, conflict among provincial courts can lead to the uneven application of federal law across provinces, whereas federal court rulings automatically have national application. Thus it stands to reason that federal officials would be more sensitive to this factor in PCA cases.

The results for winnability were mixed, as the three measures—lower court reversal, SCC ideology, and court of appeal dissent—displayed markedly different levels of success. There was no statistical support for the hypothesis, operationalized by lower court reversal, that winning at trial (or on appeal below the higher Court of AppealFootnote ¹⁷) encourages the government to appeal to the SCC. The sign of the coefficient in Table 2 was actually the opposite of that predicted, but the relationship was so far from being significant (p=0.67) that this directional finding should be disregarded. Notably, however, Reference ZornZorn's (2002:158–9) study of U.S. government lawyers found that reversal exerted an influence in the direction opposite of that hypothesized, although in his case this effect was statistically significant.Footnote ¹⁸ My findings regarding reversal are likely explained by the fact that alone of the hypothesized case factors, reversals were quite common (69 percent among cases where the federal government lost in the penultimate court), while appeals were not.Footnote ¹⁹ The ideology of the SCC appears to exercise no influence over the decision to appeal, similar to Zorn's findings for the U.S. solicitor general. This may be because the Court has been fairly consistent over time: its level of conservatism fluctuated over the 18 years studied but did not fall below 50 percent after 1986, and typically ranged between 60 and 80%. By contrast, appeal court dissent is one of the strongest explanatory factors. The federal government was 4.5 times as likely to appeal a loss from a divided appeal court as it is from a unanimous one (see Table 2). Notably, dissent was more significant for losses in the PCAs than in the FCA—in fact, it did not achieve the 0.05 significance threshold for the latter.Footnote ²⁰ This confirms the hypothesis, set out in the previous section, that winnability will be a stronger factor when a loss would expand the jurisdictional application of an unfavorable precedent.

Finally, there is no support for the rival hypothesis that the party in power influenced the rate of Charter appeals to the SCC. The Tories were actually slightly less likely to appeal (27.9 percent) than the Liberals (32.4 percent) from 1982 to 2000, contrary to expectations. This is particularly surprising given that the meaning of the Charter was more indeterminate under the earlier Tory governments. One would have thought that this would have encouraged appeals, even if the AG Canada's office is insulated from partisan influence (which is unlikely, for the reasons noted earlier). The Tories were also less likely to appeal their losses in criminal cases, this time by a wider margin—20.4 percent, to the Liberals' 33.3 percent. The hypothesis that the Tories were “tougher” in criminal appeals is, therefore, not confirmed.

Conclusion

The research presented here confirms the initial assumption of the study, that appeals by the Canadian federal government are the product of calculated decisionmaking based on costs (judicial amendment), salience (novel interpretation and intervener), and winnability (court of appeal dissent), and, in certain circumstances, reviewability (intercourt conflict). In stark contrast to the anecdotal evidence that governments routinely appeal unfavorable rulings under the Charter, the federal government is a selective appellant. This is true even among cases where the lower court has remedied legislation or crafted a new interpretive rule that is unfavorable to legislative authority generally. The evidence clearly indicates that the Government of Canada usually appeals only the most important cases—those involving novel interpretation and interveners—and those decisions constituting the deepest incursions into legislative jurisdiction, namely, judicial amendments of laws. Less-intrusive judicial remedies—including nullification of laws—surprisingly do not provoke government appeals by themselves. The simple fact of judicial activism is, therefore, less important to government decision makers than the form this activism takes, with the government challenging only the greatest incursions on its policy authority. The evidence permits the conclusion that the Justice Department actively defends the authority of the government and, correspondingly, challenges attempts by the lower appellate courts to expand judicial power, regardless of the party in power. The strong statistical results for the lower court dissent variable also support the conclusion that government decision makers actively consider strategic factors, such as minimizing or avoiding losses on appeal. To a lesser extent, strategic behavior is evidenced by the reviewability factor, as measured by the intercourt conflict variable, although the finding is robust only for appeals from the provincial courts. Moreover, recall that the categories of factors are not watertight, and the measures for winnability and salience, in particular, may also help convince the SCC to hear a case.

These findings also illustrate that strategic decisionmaking theories regarding legal actors developed in the U.S. context have international applicability, but, with the important caveats that Reference ZornZorn's (2002) work included nonconstitutional and unpublished cases and our measures differed somewhat, different factors emerged as the most influential in the two countries. Simple invalidation (“less-intrusive remedies”), interveners (on their own), fiscal costs (measured by the criminal/civil distinction), and the lower court reversal indicator of winnability did not achieve statistical significance in the Canadian context, but were four of the five statistically significant influences on the U.S. solicitor general. Notably, however, civil appeals were more likely than criminal ones in both countries, as expected, and reversals had the opposite effect in the United States of what Zorn hypothesized. Zorn's fifth significant factor—intercourt jurisprudential conflict—was only partly supported in Canada. By contrast, judicial dissent, which was among the strongest factors in Canada, was not statistically significant in Zorn's study. Thus the only factor that had comparable effects in both contexts was the SCC's ideology, which evidenced no influence. The divergent findings are quite surprising, given that the appeal decisionmaking process in both countries is centralized and at least purports to be concerned with similar factors. Future studies of Canadian government appeal decisions could be extended to nonconstitutional cases to permit a more accurate comparison with existing U.S. studies.

The government's selectivity has important implications for judicial agenda-setting. A large body of recent work on judicial activism via the “rights revolution,” led by Reference EppEpp's (1998) comparative analysis, and by Reference Morton and KnopffMorton and Knopff (2000) in Canada, emphasizes the courts' passive nature, or that judges require other actors to bring cases for adjudication. These authors and others (e.g., Reference BrodieBrodie 2002; Reference Hein, Howe and RussellHein 2001) stress the crucial role of interest groups in this regard, terming them the “support structure for legal mobilization” (Reference EppEpp 1998) or the “Court Party” (Reference Morton and KnopffMorton & Knopff 2000). This conclusion ignores that governments are by far the most frequent appellant in rights litigation. Thus the rise of the judiciary in Canada has, ironically, been made possible to a large extent by the very institution that stands to lose authority because of judicial activism. A decision by the government not to appeal losses in rights litigation, however, means that important constitutional questions are kept from being heard in the highest court in the land, or at least delayed. A notable example is whether the Charter's equality rights provisions protected sexual orientation, and accordingly whether statutory human rights codes must as well; when the government did not appeal from its loss in the lower courts on this issue, the SCC had to wait another six years to answer in the affirmative (Haig v. Canada, [1992] 94 D.L.R. [4th] 1 [Ontario Court of Appeal]; Vriend v. Alberta, [1998] 1 S.C.R. 493 [S.C.C.]).

Finally, while these findings offer an important insight into the Government of Canada's litigation behavior, it is equally important to recognize that the decision to appeal is only part of the story. Deciding to appeal to the SCC is one thing—what you argue when you get there is quite another. To date, we know very little about the purpose of government appeals; that is, what substantive goals does the government pursue on appeal to the SCC? It is to this question that future research should be directed.