Courts and Congress are, at times, engaged in a kind of ongoing “conversation” about statutory law. Congress has exclusive power to enact statutes — but when statutory language is unclear, or doesn’t explicitly resolve a factual question that arises under a statute, courts must resolve the issue through statutory interpretation. Congress then may choose to “override”1 judicial interpretations with which it disagrees (so long as the judicial decision is not constitutional in nature) by amending the law at issue or enacting a new law. The power to enact such overrides is core to maintaining democratic accountability for policy. Enactment of an override, however, is not the end of the story. As new cases arise, courts must assess how the new statutory language has changed the prior legal landscape. And so the exchange continues.
Earlier commentators, including many well-respected judges, have offered thoughtful suggestions for facilitating communication from courts to Congress about problems in statutes that Congress might want to address.2 My research explores the opposite question. How effective is communication from Congress back to courts? The answer is: Not very.3 Even when Congress enacts overrides, courts frequently continue to follow the prior judicial precedent. This is likely due more to information failure than willful disregard of controlling law. Nonetheless, a key aspect of the separation of powers is broken.
My research shows that when the Supreme Court overrules a prior decision, lower courts quickly decrease their reliance on the old precedent and begin to apply the new rule. By contrast, when Congress enacts an override, citation patterns to the prior precedent change very little. Even a decade later, many overridden precedents, or what I have called “shadow precedents,” are still routinely cited as controlling precedent. This surprising finding may be partially explained by the coding protocols used by leading legal research services. When assessing the viability of precedent, both Westlaw and Lexis consider primarily judicial signals rather than legislative signals; accordingly, it can take several years before a decision is “flagged” as having been affected by later legislation. Even when aware of an override, legal actors sometimes fail to follow the new statutory standard. Luckily, this problem is easy to address. Courts need to start their research with the statutory language itself, rather than a judicial gloss on the statutory language. Sometimes there are difficult interpretive questions regarding the scope of an override, but often it’s just a matter of carefully considering whether the operative language supersedes any aspect of a prior interpretation. By taking this straightforward approach, courts can help ensure that overrides can play their expected role in our tripartite system of government.
Congressional overrides are typically described as the legislative equivalent of a judicial overruling. My study with Professor Brian Broughman was the first to empirically test this characterization. We constructed a database of Supreme Court decisions that had been overruled by later Supreme Court decisions; Supreme Court decisions that had been overridden by later statutory amendments; and a “control” group of Supreme Court decisions that were similar (in terms of subject matter, year of decision, and other factors) to the overruled and overridden decisions but that had not been repudiated by subsequent judicial or legislative actions.4 We then used Lexis’s Shepard’s service to assess how often each Supreme Court case in our database was cited by other courts, generally looking at a 15-year window that spanned from five years prior to the superseding “event” — either overruling or overriding — to ten years after it.5 Although citation counts are admittedly a somewhat blunt measure, they are frequently used in legal and political science studies as a rough gauge of the ongoing precedential weight of a prior decision. By collecting citation data from several years before the superseding event, we were able to establish a “baseline” citation pattern, which we could then compare to citation levels after the overruling or the override. We hypothesized that citation patterns could be expected to change in two different ways: “positive” or “neutral” citations would be expected to decline, and “negative” citations, such as an indication that the prior decision had been fully or partially overruled or superseded, would be expected to increase. To capture both of these effects, we developed a measure we called “net citations,” which we defined as the number of positive or neutral citations to a decision, minus the number of warning or other negative citations.6 We then compared the average number of net citations a case received each year after the event to the average number of net citations the case received before the event; this ratio measures how much effect the overruling or override had on citation levels.
Our findings were striking. As shown in Figure 1, after a judicial overruling, net citations to the prior decision drop rapidly when compared to the pre-event baseline. The citation patterns for cases in our “overridden” category, by contrast, are very similar to those of our control group. Overall levels of citations drop, but in a gradual fashion that is typical of the natural “depreciation” that decisions generally experience over time.7 Even ten years after an override is enacted, most overridden precedents are still widely cited as controlling precedent.
Degree of Overruling or Override. We recognize that an override may supersede some, but not all, of the analysis in a prior decision, meaning other aspects of the decision remain controlling. The same, of course, is true for a judicial overruling. To assess whether this affected our results, the cases were assigned a “depth” measure that evaluated how completely the overruling Supreme Court decision or overriding legislation rejected the prior opinion,8 as well as an “explicitness” measure that evaluated how explicit the Court or Congress was about its disapproval of the prior opinion. We found that for both sets of cases, greater “depth” was associated with a larger decline in citations; however, at each level of “depth,” citations to overruled cases declined more dramatically than citations to the overridden cases. The same was true for “explicitness.” Thus, our findings are not the result of comparing deep and explicit overrulings to shallow and non-explicit overrides. Rather, even when we control for these factors, we find that judicial overrulings have considerably more effect on future citations than legislative overrides.
As an additional robustness check, for a randomly selected subset of cases in both groups, we hand-coded individual headnotes to distinguish between headnotes identifying portions of the prior decision that had been superseded and those that had not. Since Lexis’s Shepard’s service tracks citations to each headnote in a case, this allowed us to assess in a more fine-grained manner which propositions within each case were being referenced when later decisions cited to the earlier precedents. For both groups of cases, we found a notable decline in net citations to the headnotes associated with specific propositions within the cases that had been superseded, but again this decrease was much more pronounced for the overruled cases than the overridden cases. Additionally, we assessed the extent to which ideological preferences might explain ongoing citation of overridden precedents, but our data did not suggest a judge’s ideology was the driving factor.9
Prospectivity. Because a judicial overruling is a reinterpretation of existing law, it typically takes effect immediately; the Court’s new interpretation will apply to all pending disputes, including those arising out of events that pre-dated the new opinion. By contrast, statutory overrides are typically prospective; the old (now superseded) judicial standard will govern the resolution of a dispute arising out of events that pre-date the effective date of the statutory amendment, even if the decision in the case is issued after the effective date of the amendment. For this reason, we would expect to see a judicial overruling have a more immediate effect on net citation levels than a statutory override. To address this issue, our analysis excluded citation counts from the year of the superseding event and the first two years after the superseding event, as this is the window when we expect the difference between retroactive judicial overrulings and prospective statutory overrides to be most salient. We modified these parameters to exclude greater and fewer years, but our general results held, suggesting that the differences we observe are not driven by the prospective nature of overrides.10
Since this study relies on citation counts, rather than a close reading of the context for each citation, we cannot definitively assert that any particular citation of an overridden case was in error. Below, I provide specific examples, drawn from my work on overrides in the employment discrimination context, of both “proper” and “improper” citations to overridden cases. The big picture conclusion is clear, however. If overrides were having the effect that they are intended to have, it is reasonable to assume that there would be sizeable decline in citations to legislatively overridden precedents, just as there is a sizeable decline in citations to judicially overruled precedents. Instead, on average, citation patterns to the overridden cases are almost indistinguishable from those to the comparison control group of cases that have been neither overridden nor overruled. This suggests that often courts fail to hear — or to heed — Congress’s side of the dialogue.
Judicial overrulings may be more effective than legislative overrides in part because the coding protocols used by leading legal research services, such as Westlaw’s KeyCite system and Lexis’s Shepard’s system, are far better at identifying the former than the latter. The services’ coding protocols rely almost exclusively on judicial signals rather than legislative signals.11 This approach reflects the common law roots of American law; the coding protocols are based on the assumption that only a subsequent judicial decision will affect the precedential weight of a prior judicial decision. Under these protocols, when the Supreme Court overrules a prior precedent, coders will generally add a “red-flag” or a “warning” signal to the prior precedent immediately. By contrast, coders are generally not expected to assess whether legislative actions affect the precedential value of judicial decisions; an overridden precedent typically will not be flagged unless or until a court asserts in a decision that new statutory language has superseded a prior judicial interpretation. After a court has made such a declaration, coders should add a cautionary or warning signal to the prior decision. However, Westlaw will only red-flag a decision if a court at the same or a higher level in the judicial hierarchy indicates a change in the precedential value of the case. This means that Supreme Court cases that have been overridden will only be red-flagged on Westlaw if the Supreme Court itself explicitly indicates that the prior decision has been superseded by the subsequent legislation.12
This can result in surprisingly long lag times — on average, more than four years (!) — before a legislatively overridden precedent is flagged at all, and only about 20 percent of overridden Supreme Court decisions are red-flagged on Westlaw.13 That said, this average masks significant differences. A leading study distinguishes between overrides that explicitly denounce prior judicial interpretations as incorrect, what the study terms “restorative” overrides, and overrides designed to update or clarify a statute, often in response to a judicial invitation to do so.14 Many precedents superseded by restorative overrides are flagged within a few months. By contrast, it often takes several years before precedents that are affected by an updating or clarifying override are flagged; some may never be flagged. This may be because updating or clarifying overrides receive less attention from popular and legal media. They are also often part of larger overhauls of statutory law, such as a major reform of bankruptcy or tax law, rather than standalone bills that are clearly responsive to prior Supreme Court decisions. The irony is that courts often explicitly invite Congress to enact updating or clarifying overrides, on the theory that new policy should come from the legislature rather than the judiciary — but even if Congress does so, the old precedents may continue to hold sway.
My work in this area has focused on congressional overrides of Supreme Court decisions. However, Congress also frequently supersedes lower court decisions, sometimes resolving a circuit split or an area of confusion before the Supreme Court weighs in.15 While my research did not look at this directly, it is likely that there may be even longer delays before such lower court decisions are flagged as superseded.
My empirical study with Professor Broughman shows that citations to overruled precedents decline quickly, but citations to overridden precedents, on average, do not. This suggests that overrides do not have as much impact as they should. In part, this is likely due to the Westlaw and Lexis coding protocols discussed above. But even when aware that an override has been enacted, lower courts must assess the extent to which the new statutory language supersedes the prior judicial interpretation. This is often straight-forward; it simply requires that courts carefully apply the standard in the revised legislation. But in some cases, it can be difficult to determine the scope of the override. I have conducted case studies of employment discrimination overrides that illustrate some of the issues that can arise.
Mistakes in Applying New Statutory Language. The ADA Amendments Act of 2008 (ADAAA) explicitly repudiated prior Supreme Court decisions that had interpreted the meaning of “disability” in the Americans with Disabilities Act very stringently.16 The ADAAA’s stated purpose was to “reinstat[e] a broad scope of protection” under the ADA.17 It passed with high levels of bipartisan support and was signed by President George W. Bush. Several years after it was enacted, I assessed the degree to which courts were appropriately following the new statutory standards.18 I found that, with some regularity, courts continued to apply the overridden cases for propositions that were unquestionably superseded by the new statute.19 The mistakes fell into two categories. First, there were decisions that failed to mention the ADAAA at all, even though the events giving rise to the dispute occurred after the effective date of the ADAAA.20 These cases simply applied the prior precedent as if the ADAAA had never been enacted. Second, there were many decisions in which courts observed (correctly) that the ADAAA had superseded the pre-ADAAA precedent in some respects but then asserted (incorrectly) that the relevant portion of such precedent still applied, even in contexts where the judicial interpretations at issue were clearly and indisputably incompatible with the new statutory language. Courts also sometimes observed (correctly) that the ADAAA had superseded pre-ADAAA Supreme Court precedent, but then followed circuit court precedent based on the superseded Supreme Court precedent, without acknowledging that the circuit precedent should also be reconsidered in light of the new statutory language. For example, numerous lower courts applied prior Supreme Court and circuit court precedent concerning what was necessary to establish that a plaintiff was “regarded as” having a disability, even though the amended statute established a new, far more flexible standard for “regarded as” claims.21 Our adversarial system depends primarily on lawyers to identify the relevant law and make arguments that will benefit their clients. I assume that in many of these cases, the lawyers failed to properly use the new statutory standard in their briefs. However, even if lawyers make such mistakes, judges have an independent responsibility to apply the governing statutory law.22 Additionally, sometimes the party who would have benefited from the new language was pro se, making it even more imperative that judges research the applicable law carefully.
In many instances, the legal mistake may not have affected the outcome of the specific case, as there may have been independent grounds that justified the ultimate decision. However, the error is still problematic. I found several cases in which the flawed reasoning of one court was cited and followed by other courts. In other words, once a court has incorrectly characterized the scope of an override, it can be difficult to stamp out.
Moreover, the ADAAA was an unusually direct override that received significant attention in both the popular media and the legal press. Indeed, in terms of absolute numbers, citations to the precedents overridden by the ADAAA dropped quickly and dramatically.23 The prevalence of mistakes even in this high-profile context suggests that mistakes likely make up a large portion of ongoing citations to overridden precedents in general, since most overrides are not as clear or prominent as was the ADAAA.
Interpretative Complexity Sometimes Arises in Implementing Overrides. I have also explored situations in which judges could reasonably disagree about whether the override statute or the pre-existing precedent should apply. One common question is whether the statutory language fully repudiates the reasoning, as well as the result, of a prior decision. For example, in 1978, Congress overrode a Supreme Court decision that had held that pregnancy discrimination was not a form of sex discrimination.24 Lower courts have since disagreed about how to resolve cases involving similar issues, such as discrimination on the basis of breastfeeding or contraceptive access. Some courts have concluded that Congress implicitly adopted the reasoning of the dissenting justices in the original case, whereas other courts suggest that the majority’s reasoning continues to control resolution of these related matters.25
It may also be unclear how an override should be interpreted when Congress amends one statute to supersede a judicial interpretation but does not make comparable amendments in other statutes that are typically interpreted in pari materia. For example, in 1991, Congress amended Title VII of the Civil Rights Act to partially codify and partially override a judicially-crafted causation standard. Congress did not amend other employment discrimination statutes, but the House Judiciary Committee stated in a report that it expected these statutes would be interpreted “consistently” with the standard in the override.26 When deciding cases brought under other discrimination statutes after the 1991 amendments, some lower courts applied the causation standard in the override, and others applied the prior precedent.27 Ultimately, the Supreme Court rejected both these standards, announcing a different standard that it held applied to the private-sector provisions of the age discrimination statute, and, in a subsequent case, Title VII’s retaliation provisions.28 Courts at all levels continue to grapple with how the Supreme Court’s more recent decisions apply to other federal statutes.29 As this article was being finalized for publication, the Supreme Court was actively considering two more cases that arise from this series of events.30 Courts must also determine what causation standard applies to state laws modeled on these federal laws, a question made particularly difficult by the fact that many state statutes address age along with traits covered by Title VII.31 Similar questions arise even when Congress agrees with a judicial interpretation, if it codifies that interpretation in the statute but does not make comparable changes in other statutes with similar language.32
In my earlier work, I argue that relying on overridden precedents in these kinds of situations can undermine interests usually served by following precedent, such as fairness, efficiency, and predictability. Such reliance can also thwart congressional intent. In the causation cases discussed above, the Supreme Court has suggested that if, in 1991, when Congress amended Title VII to supersede the Court’s prior interpretation of Title VII, Congress wanted its preferred causation standard to apply more generally, it should have simultaneously amended all other anti-discrimination statutes with language similar to that in Title VII. The Court then characterized Congress’s failure to do so as an affirmative “choice” by Congress that a different standard should apply.33 I have argued that this conjecture does not reasonably reflect the institutional realities within which Congress operates. As an alternative, I have suggested interpretive conventions that I believe would allay confusion about the scope of overrides and better accord with likely congressional intent. Specifically, I argue that enactment of an override should create rebuttable presumptions that (1) both the reasoning and result of the prior interpretation is superseded; and (2) similar language in other statutes should be interpreted in accordance with the override, so long as that application is a reasonable interpretation of the pre-existing statutory language.34 A full explication of these proposals is beyond the scope of this article. Here, I simply highlight that there are sometimes complicated interpretative questions implicit in interpreting overrides. This may be at least a partial explanation for why overrides have less effect on subsequent citation patterns than overrulings. To address this problem, Congress or the Supreme Court should provide greater clarity about how lower courts should resolve these kinds of questions.
Overrides are not self-implementing. They are only effective if other legal actors properly apply the new statutory standard, rather than the prior judicial precedent. As this article has shown, that process often breaks down. Luckily, there are some ready fixes.
Start with the Statute. When doing legal research in a statutory case, lawyers and courts should always begin their analysis with the statute itself. If judicial precedents interpreting the statute predate any statutory amendments, the current operative statutory language should be carefully assessed to determine whether and to what extent it supersedes the prior judicial analysis. It may be helpful to consult the “finding and purposes” sections for the statute, as Congress may identify particular judicial decisions that have motivated statutory changes.35 Such preambles may not be codified adjacent to the substantive provision, or they may not be codified at all, but they can easily be found in the public law. Committee reports and other reliable legislative history can also provide helpful context.
Explicitly Recognize Overrides in Opinions. As discussed above, the coding protocols used by Westlaw and Lexis rely almost exclusively on courts to assess the precedential weight of prior decisions. This means there can be a multi-year delay before an overridden precedent is flagged as superseded, and some overridden precedents are never flagged. Judges can mitigate this problem simply by stating explicitly in their decisions that a statutory amendment supersedes or partially supersedes a prior precedent. Such statements should result in the earlier decision being flagged, even if the court ultimately determines it is appropriate to rely on the overridden precedent in resolving the particular question at hand. In subsequent disputes, lawyers and courts will then be more likely to assess properly whether the new statutory language affects the application of the prior precedent.
Identify Overridden Precedents in Statutory Language. Congress should also take steps to improve the efficacy of overrides. Congress often provides explanations for why a statute is being amended, including approval or disapproval of judicial decisions, in committee reports.36 When possible, comparable statements should be included in the statute itself. This should facilitate more prompt flagging by legal research services.37 Additionally, this would likely be more effective in informing statutory interpretation, as there are some judges who categorically refuse to consider legislative history. However, large-scale revisions of substantive law often implicate numerous Supreme Court and lower court decisions; it could be quite unwieldy (and counter to existing drafting norms) to mention all relevant decisions in the statutory text. Accordingly, Congress’s failure to identify a precedent as impacted by a new law should not be interpreted as a signal that the prior precedent remains controlling.38 If the substantive statutory language does not accord with the interpretation in a prior judicial opinion, the statutory language controls, whether or not the statute mentions the prior decision.
Clarify Effect on Related Statutes. When enacting an override, Congress should also consider whether there are other statutes that are typically interpreted in pari materia and, if so, whether it intends to have the override apply to these other statutes. Where practicable, such as where there is a clearly defined and limited number of statutes with similar language, Congress could make changes to all such statutes.39 In some instances, it might be extremely difficult to identify and amend all other statutes with similar language.40 Nonetheless, Congress should at least make statements in “findings and purposes” clauses that indicate its intent as to the scope and application of an override. (These same approaches also apply when Congress agrees with or essentially codifies a judicial interpretation.) I have also argued the Supreme Court should reconsider the inferences it draws from congressional “silence” in this context to better reflect the realities of congressional process.41
Improve Legal Research Functionality. Finally, Westlaw and Lexis should consider making changes to their coding protocols to identify overridden precedents more clearly and more quickly. The Congressional Research Service, the Legislative Counsel’s office, or other relevant agencies also could compile lists of overrides. Any and all of these reforms can help overrides better achieve their intended effect.
My research in this area boils down to a simple point: Overrides often fail to actually override. By taking steps to ensure that statutory research properly assesses the significance of legislative change, and by recognizing that there may be significant lag time before legal research services flag earlier decisions as having been superseded, courts and lawyers can play a key role in making overrides more effective. This will protect and promote core principles of democratic accountability built into the structure of our government.