Judicial Sunk Cost Bias I. Introduction This paper hinges upon the assumption that judges are vulnerable to a sunk-cost bias, i.e., they decline to overrule legal decisions that were heavily invested with prior judicial resources. The sunk cost bias is an irrational bias that is likely to entail sub-optimal decisions. Hadn't they been vulnerable to the bias, the argument goes, judges would have decided differently. The application of the sunk cost insight to the judicial arena is, in itself, quite novel. 1 Its novelty notwithstanding, this simple application has a bearing to any area of law in which a sequential decision-making is required (such as the following phenomena: overruling precedents, new trials and habeas corpus writs, judges sticking to prior determinations upon a remand, alleging trump arguments, such as recusal, as close as possible to the outset of the trial) This paper attempts to concoct an empirical examination of the judicial sunk cost bias. II. Theoretical and legal framework The sunk cost bias, formally known as the escalation of commitment problem, is also recognized in the organizational context, where two sequential decisions 1 One commentator appears to draw a similar assumption while claiming that the stare decisis tradition reflects a sunk-cost bias (Rafael Gely, Of Sinking and Escalating: A (Somewhat) New Look at Stare Decisis, 60 U. Pitt. L. Rev. 89 (1998)). Nonetheless, he neither proceeds to other areas of law nor tries to prove the underlying assumption empirically. The sunk cost bias is by no means unknown to the legal literature, though in other, non-judicial, contexts (see, for instance, Samuel Issacharoff and George Loewenstein, Second Thoughts About Summary Judgment, 100 YALE L. J. 73, 113-114 (1990) (raising the argument that the sunk cost bias impedes litigants who invest considerable money in litigation from settling)). 1
are not necessarily taken by the same organ. There are various psychological and sociological explanations to the sunk cost bias. One important application of this bias concerns the odds of reversal: the more a trial judge invests in a case, according to this reasoning, the appellate court would be less likely to reverse the decision. Meritorious appeals are, therefore, unduly denied. Interlocutory appeals, then, have a remedial effect on the sunk cost bias. Sunk cost considerations should tilt towards choosing a liberal rule of interlocutory appeal ( piecemeal litigation, typical to New York and Israeli courts, rather than the final judgment rule, characteristic to federal courts). Interestingly, and according to the sunk cost logic, current federal rules of civil procedure endorse the so-called death-knell theory to permit a more liberal interlocutory appeal on class action certification decisions. III. The difficulties in observing the sunk cost bias There are at least three different ways to strengthen the assumption of judicial sunk cost bias. (a). Observing appellate decisions that hint that judges do take into account the significant (and probably non-duplicable) judicial resources that were expended at the trial court. 2 (b). Experimental study that would attempt to gauge the judicial sunk cost bias by creating laboratory conditions. (c). Empirical method: finding a correlation between reversal rate and sunk judicial resources at the trial court. This paper takes the latter approach. A correlation between reversal rate and judicial resources, however, may reflect at least two alternative phenomena. (a). Larger judicial resources at the trial level 2 Motorola Credit Corporation v. Uzan, 388 F.3d 39, 55-56 (2 nd Cir., 2004). 2
indicate that the court had a better factual acquaintance with the case. Unwilling to meticulously examine the record, appellate judges would take larger judicial input as a proxy for a decent factual inquiry. (b). Similarly, a judge who invests more resources in a certain case is deemed to generate a better legal decision. In order to overcome these pitfalls as much as possible, I suggest focusing on jurisdictional issues. Typically, questions of jurisdiction are pure legal ones, and do not require a factual inquiry. Indeed, those issues should be decided at the outset of the litigation, before any evidence was proffered. The trial court, thus, does not have any relative advantage vis-à-vis the appellate instance. In addition, lack of jurisdiction is too crucial an issue to rely upon ambiguous legal determinations of the trial court. Without jurisdiction the court cannot proceed... when it ceases to exist, the only function remaining to the court is...dismissing the cause. 3 Appellate judges are specifically directed to ignore the signals sent by the trial court and review these questions de novo, as opposed to the ordinary, laxer, standard of review. Theoretically at least, a want of jurisdiction is never a harmless error. Considerations like the amount of work invested in this very issue or in subsequent litigation should not matter to the appellate instance even if the substantive result is totally correct. Indeed, a challenge to jurisdiction can be raised at any time, even at the appellate instance. 4 In addition, appellate courts are obliged to raise jurisdictional difficulties sua sponte. Policy considerations aside, these strict rules of jurisdiction are well-entrenched in the American legal history and the common law tradition. Finding a correlation between the reversal 3 Ex parte McCardle, 7 Wall. 506, 514, 19 L.Ed. 264 (1868) 4 Capron v. Van Noorden, 2 Cranch 126, 127, 2, L.Ed. 229 (1804) 3
rate of jurisdictional issues and the amount of judicial resources that were invested at the trial court can hint, therefore, at judicial sunk cost bias. This is the underlying hypothesis of my research. IV. The methodology I looked for a general correlation, within the Federal system, between the reversal rate of jurisdictional issues and judicial input invested by the trial court. Three different factors were used to reflect judicial input in the trial court: 1. The length (pages) of the decision/s rendered at the district court; 2. The period (months) in which the case appeared at the trial court's docket; 3. The procedural stage in which the case was terminated (i.e., motion to dismiss, summary judgment, jury trial etc.). In addition to signifying judicial input, these three parameters should not imply that the jurisdictional question was rightly resolved. 1. Conceptual difficulties of the methodology. There are two preliminary difficulties: firstly, the Federal legal system uses a different organ to decide the subsequent decision, and employs elite American lawyers as decision-makers. Thus, it may suffer a negligible sunk-cost effect. The second set of problems, which was discussed, concerns the efficacy of the reversal rate of jurisdictional issues to serve as a benchmark to the judicial sunk cost bias. Thirdly, there may be an endogenous adjustment effect, namely, the system's ability to adapt itself to the bias and eliminate its apparent outcomes. At least two possible alterations, then, may remove the sunk cost bias effect from the findings. 4
(a). Litigants' behavior changes. Anticipating the sunk cost bias effect on appellate courts, litigants may adapt by: 1. Not appealing on jurisdictional questions when those were followed by a significant judicial input. 2. Find creative ways to an immediate appeal where interlocutory appeal is not formally allowed. However, it is by no means self evident to assume that litigants are rational enough to sense judges irrationality. (b). Judges' behavior changes. In a similar vein, trial judges might change their behavior in response to the influence of the sunk cost bias on appellate judges. Driven to maximize their leisure and lower their odds to be reversed, trial judges may lower the quality of their jurisdictional determination if extensive litigation is likely to follow. Again, this argument is by no means self-evident: one needs to assume that trial judges are aware of appellate judges' irrationality. V. The database I used online resources (Westlaw) to collect all the relevant cases. The final database consists of 75 Second Circuit 2003-2007 cases in which a substantial question of jurisdiction was upheld by the district court. When I picked cases to the database, I relied upon the circuit's court definitions for jurisdiction. Thus, my sample includes all types of jurisdictional issues. VI. Results and discussion I found that two parameters, the length of the first decision and the time it was pending, have no sunk cost effect. More interesting is the third parameter the procedural posture in which the district court concluded the case. The sample can be divided to two groups: the first consists of 66 cases in which a jurisdictional 5
question was upheld and the case was eventually decided in an early procedural stage, without a jury or bench trial (i.e., by a motion to dismiss/summary judgment). The odds for reversing the jurisdictional decision in this group are 21.2% (14 out of 66 cases), a plausible number. The second group consists of 7 cases in which a jurisdictional question was raised, upheld, and then the case proceeded to a full trial. In this group (after a full trial) the reversal rate is 0% (0 out of 7): Decision rendered by the lower court before completion of a bench or a Decision rendered by the lower court upon completion of a full trial jury trial Jurisdictional issue reversed 14 (21.2%) 0 (0%) by appellate court Jurisdictional issue affirmed 52 (78.8%) 7 (100%) by appellate court Total 66 (100%) 7 (100%) The results are consistent with my logic. Nonetheless, the second group cases that were appealed after the district court upheld its jurisdiction and then continued to a full trial consists of remarkably low number of cases. It might suggest that litigants are aware of the sunk cost bias and somehow avoid its consequences. 6
VII. Conclusions and further research The empirical findings. A further research is needed. It can draw on the same methodology, significantly enlarging the database. Plus, it should focus on subject-matter jurisdiction a clearer type of non-factual and non-waivable jurisdiction. I am working on this further research in these days. Prescriptive implications. Firstly, some modifications can be implemented in order to enable judges to better cope with the judicial sunk cost bias. For instance, judges may be required to provide reasoning. A second set of resolutions aims at avoiding, ex-ante, any hazard of sunk cost bias. This logic entails a broader right to interlocutory appeals. Thirdly, the troubles that are attributed to the sunk cost bias can be neutralized by altering the substantive law. A more liberal standard of habeas corpus and new trials, for example, may counter affect the judicial sunk cost bias. 7