|Volume 75||Number 4||Winter 1997|
We have three things to think about here, as the real estate agents say--"location, location, location." Accordingly, the two of us have engaged for several years in empirical studies aimed at gauging the effect of forum on case outcome. The results to date strongly suggest that forum really matters.
An early piece of the puzzle fell into place in our study on transfer of venue. In that article, we examined the benefits and costs of the federal courts' scheme of transfer of civil venue "in the interest of justice." Ours was a pretty straightforward and simple cost-benefit analysis, but we supported it with some fairly elaborate data. For example, to expose the benefits, we demonstrated a dramatic drop in plaintiffs' rate of winning after transfer of venue, and we argued that on average the altered outcomes reflect the more just transferee forums' more accurate outcomes. Thus, the benefit side of the analysis included this substantial reduction in error costs, along with direct costs' net diminution from litigating in a more convenient forum. As to the cost side of the analysis, our empirical work indicated considerable total expense of handling transfer motions, but also a tendency among commentators to overestimate that expense. Our bottom line, admittedly not definitive, was, "[g]ood policy calls, at the least, for preserving the transfer mechanism."
Now Professor David Steinberg, after some kind introductory words, responds, "I disagree with almost all of their conclusions." He specifies:
First, by including cases where courts have entered default judgments, Professor Clermont and Professor Eisenberg exaggerate the effect, if any, that the choice of forum has on the outcome of a case. Second, Professor Clermont and Professor Eisenberg do not demonstrate that transfers lead to more accurate outcomes. Instead, their data only supports the conclusion that a defendant's chance of winning a case improves if the defendant has selected the forum through a transfer motion. Third, Professor Clermont and Professor Eisenberg understate the costs and appropriateness of transfer under the open-ended standard currently employed in section 1404(a) litigation. Fourth, if Professor Clermont and Professor Eisenberg are correct that forum shopping is both pervasive and effects outcomes, Congress or the courts should address this problem directly by limiting the geographic choices available to a plaintiff bringing a federal court suit. Such an approach would address any inequities resulting from forum shopping far more universally and efficiently than the case-by-case transfers currently employed under section 1404(a).
Professor Steinberg's first two arguments attack our benefit analysis, as he contends that we failed to show any real drop in plaintiffs' win rate upon transfer and that, even if we did succeed, we have no reason to suspect that the change is in the direction of accuracy. His third argument goes to our cost analysis, as he contends that the expense of handling transfer motions is exorbitant. Professor Steinberg's fourth argument addresses our conclusion, as he says that if against all odds we were right, then we should be championing the more radical reform of narrowing the plaintiffs' initial choice of forum.
We believe that our article met Professor Steinberg's first three arguments. As to his fourth point, however, we are in agreement. Our point had indeed been to open the door to that more radical reform, but we gently pushed it open because we were early in the process of gathering empirical support. Since writing that article, we have extended our focus beyond transfer. We have shown that forum affects outcome in other contexts. Therefore, we are now prepared to argue ever more explicitly that the initial choice of forum is currently a little too wide to be just. At any rate, we shall reply to Professor Steinberg's four arguments in order.
The striking finding that triggered our research was that in three million recent federal civil cases, the plaintiffs won 58% of the non-transferred cases that went to judgment but only 29% of such transferred cases. One cannot look just at output, however, because the set of non-transferred cases is very different from the set of transferred cases. For example, we noted prominently that about 32% of non-transfer judgments are by default, contrasted to about 4% of transfer judgments. In other words, "easy plaintiff victories often, but not always, occur quickly, without transfer, so that transferred cases are tougher to win." Accordingly, we warned right at the outset that "the overall drop from 58% to 29% gives an exaggerated sense of the magnitude of the transfer effect."
The exaggeration was not a wild one, however. Just as we noted several sources of exaggeration, we noted several sources of understatement. Most notably, we were looking only at judgments. Yet 53% of non-transfer case terminations are by non-judgment dismissal (primarily voluntary dismissals and dismissals for lack of prosecution), while 64% of transfer terminations are by non-judgment dismissal. This contrast shows that plaintiffs tend to abandon their cases after transfer. Therefore, confining our study to judgments ignored this abandonment and thus understated the actual transfer effect.
Nevertheless, to remain as objective as possible, we did restrict our focus to judgments for the plaintiff or the defendant. We then employed the mathematical technique of regression to control for the differences between the non-transfer and transfer judgment sets, including the difference in default rate. In other words, we separated out the effects of the differences between cases from the effect of transfer by itself. We found that "transfer continues to have a substantial negative effect on win rate," so much that transfer reduces a 50% chance of the plaintiff's winning a judgment to 40%.
Interestingly, we have since expanded our footnoted comparison of transfer and removal into an article on the "removal effect." There, looking at judgments for the plaintiff or the defendant again, we conclude that removal by itself reduces a 50% chance of plaintiff's winning to 39%. This comparison suggests a consistent "forum effect" that reflects the plaintiff's loss of forum advantage.
Very often empirical results are counterintuitive. Most readers are consequently loath to accept them. Yet here our work substantiated common knowledge. As most practitioners already knew, forum-shopping is the name of the game. Nonetheless, Professor Steinberg remains unconvinced that forum in fact affects outcome. Professor Steinberg dismisses our fifteen pages of data analysis simply because the "28% difference in default judgment rates is almost identical to the overall 29% difference in plaintiff win rates for non-transferred as opposed to transferred cases."
Because empirical investigation and interpretation are so difficult, those who dislike the results will not find it this easy to dismiss or explain away those results. First, according to the data presented in our article, the plaintiffs' win rates in non-transferred and transferred cases, excluding default judgments, are 41% and 27% respectively. Our discussion of methods of case disposition, for example, noted for judgments by pretrial motion a non-transfer win rate of 26% and a transfer win rate of 13%. Second, Professor Steinberg's observation indicates a lack of understanding of regression analysis. Quite simply, we took the different default rates into account, at some length in fact. The regressions proved the persistence of the transfer effect. Third, even at a lower level of mathematics, Professor Steinberg's observation ignores the rules of ratios, and the changing denominator. If one excludes default judgments at the rates of 32% and 4% and assumes they were all plaintiff victories, the 58% non-transfer and 29% transfer win rates do not fall to equality as Professor Steinberg implies, but after a simple calculation stand at 38% for non-transferred cases and 26% for transferred cases.
In summary, the plunge in default rate after transfer by no means fully accounts for the observed transfer effect. The effect of transfer in reducing the plaintiffs' win rate is real. Unsurprisingly, the plaintiffs have an advantage in the forum chosen by the plaintiffs. Indeed, instead of detracting from our position, the plunging default rate after transfer may support it. The high default rate in the plaintiffs' initially chosen forum may in part reflect their real forum advantage, as some defendants just elect not to litigate there.
Perhaps we need to make a more general point, one that goes beyond defaults and mathematics. Why is it that Professor Steinberg and others persist in thinking that forum has a negligible effect in comparison to the merits, while our research seems to prove that forum can have a serious effect on outcome? A clue lies in one of Professor Steinberg's early footnotes:
Professor Clermont and Professor Eisenberg do not discuss the radical implications of their suggestion that transfers lead to more accurate outcomes. . . . But if the 29% plaintiff win rate in transferred cases is accurate or just, then presumably the 58% win rate in cases that courts do not transfer is inaccurate or unjust. Following this reasoning, courts should consider transferring every case.
Although otherwise faulty, Professor Steinberg's point shows that he is thinking of all cases, while we are observing the transfer effect only in the small subset of transferred cases. As we wrote:
One would not expect the forum to have a mammoth effect independent of the merits. However, transferred cases do tend to comprise those cases where the substantive effect of forum is greatest. After all, the plaintiff tried to forum-shop, the defendant chose to fight back, and the court decided that the forum truly mattered. Thus, it is reasonable to conclude that redressing plaintiffs' forum-shopping advantage by transfer should decrease the win rate to some extent.
In other words, we are not saying that forum-shopping always has a major impact. In most cases the plaintiff has chosen a basically fair forum. We are simply saying that in the relatively few cases that the courts transfer, the transfer procedure has offset a serious forum effect.
Thus forum, and hence transfer, affects outcome. But in what way? On this we argued that:
as transfer works to neutralize any lopsided cost advantage, and thereby to equalize the effectiveness of the two sides' litigation expenditures, the outcome should be more accurate in the transferee court. . . . Transfer removes the plaintiff's forum advantage when the interest of justice so counsels, and therefore removes the plaintiff's opportunity to gain an unjust victory in litigation or to achieve an unjust settlement.
Professor Steinberg attacks here too, again with a single line of argument: because transfer shifts the case to a forum chosen by the defendant, and forum-shopping by the defendant is no more conducive to accurate outcomes than forum-shopping by the plaintiff, we cannot draw any normative inference from changed outcomes.
This common view misses how transfer of venue works. The plaintiff initially chooses the most favorable forum by filing suit in that one court off the list of proper venues. If the choice is too favorable to tolerate, the defendant moves to transfer to another court off that same list. The judge, typically more neutral and objective than the parties, orders transfer only if it is in the interest of justice. The key difference between the transferor court and the transferee court is that the plaintiff unilaterally chose the former, while the defendant must get the judge to agree that the latter is not only a better forum but a much better forum.
Because the transfer process is far from unrestrained forum-shopping by the defendant, Professor Steinberg's argument falls apart. We accordingly summarized that:
transfer does not shift the choice of forum from plaintiff to defendant, but instead from plaintiff to judge. Moreover, the judge decides to transfer only in rather extreme cases of forum-shopping, normally deferring to the presumption in favor of the selected forum. In short, the transferee forum should generally be a better forum affording a better outcome.
Critics of transfer without any data tend to assume very high costs of operating the transfer system, and therefore we set out to provide some numbers. We concluded that transfer motions occur in less than 5% of cases, and that most of these motions experience inexpensive handling in the district court and no appeal. Some of the motions have a clear result, reducing the effort that goes into them. Even on most of the closer ones, despite the importance of forum to the parties, the court and the litigants cannot sink grand amounts of resources into the venue fight because of the inherent simplicity of the issues.
Professor Steinberg responds only that all this expense still sounds like a lot to him. Well, it sounds fairly considerable to us too, but not in comparison to the benefits of the transfer system.
Professor Steinberg continues, however, by taking this opportunity to resurrect his proposal to change the standard under the federal transfer statute, a position to which he committed himself in an earlier article. Professor Steinberg argues that to reduce costs, courts should decide transfer motions usually by considering only the location of the preponderance of relevant witnesses and documents. We certainly would not resist any sound attempt to reduce transfer's costs. However, Professor Steinberg's proposed narrowing of judicial focus to one factor does not seem remotely to capture the appropriate standard of "in the interest of justice." Moreover, Professor Steinberg's cramped focus will not always produce easy questions for judicial decision. It could even induce more motions to transfer because defendants could often make a plausible argument for transfer based only on a preponderance of relevant witnesses and documents. It indeed cannot, and does not, completely close the door to the litigants' arguing or the courts' considering other factors, such as the personal circumstances of the parties. Finally, Professor Steinberg's proposal would have profound but unforeseen effects, especially upon plaintiffs, as it is almost equivalent to amending the venue statute to require plaintiffs to bring every suit in the one court with the preponderance of relevant witnesses and documents. In sum, the matter of which forum will be better in producing the better outcome turns on many factors, and any standard that arbitrarily restricts the courts' focus to one factor is ill advised.
Having established that the various benefits of transfer outweigh the costs of administering transfer, we concluded that the legal system, "at the least," should preserve the transfer mechanism. Here Professor Steinberg, for the sake of argument, switches sides to accuse us of timidity.
However, we expressly observed that there was nothing wrong with a proposal to tighten the plaintiffs' initial choice of forum. One of us, moreover, had already written in favor of "a progressively more restrictive approach on both federal and state levels" to this question of territorial authority to adjudicate and, more to the point and on the basis of our transfer article, subsequently made the following proposal:
So plaintiffs today are frequently forum-shopping their way to unjust victories. A role therefore remains for due process--as a robust test of multifactored reasonableness, not a narrow test limited to extreme inconvenience or bias effectively denying a day in court or limited to a few factors such as unfair choice of law [as some have suggested]. We must rely on due process to ensure that the plaintiff is choosing from a limited list of basically fair forums, while also relying on venue provisions to narrow further the choice within a workable system and relying on transfer mechanisms to handle the special cases.
At the time of our transfer article, we were not far enough along in our empirical research to do more than suggest such radical reform. Now that we have studied the forum effect in settings other than transfer of venue, we are ready to argue that the plaintiffs' initial choice of forum is currently a little too wide. Because states have incentives to extend their territorial reach, the most plausible method for restricting them is federal judicial control, which would entail a somewhat greater role for due process. On the federal level, we can hope and argue for a tighter venue statute from Congress. Thus, one could aspire to the situation that Professor Steinberg professedly prefers: "Parties should win or lose cases based on the merits of their substantive claims, and not as a result of procedural maneuvering." Yet Professor Steinberg seems all too content to preserve the current system, where the outcome sometimes turns on the plaintiff's choice of forum.
Professor Steinberg also contends that the Court's and Congress's spotty performances to date in the jurisdiction and venue realm somehow establish that current regulation of the plaintiffs' initial choice of forum is optimal. With the new information provided by our empirical research, we are trying to change that current law and current thinking. Professor Steinberg's reaction indicates that ours will be an uphill battle.
[*] Flanagan Professor of Law, Cornell University.
[**] Henry Allen Mark Professor of Law, Cornell University.
[1.] See Kevin M. Clermont & Theodore Eisenberg, Exorcising the Evil of Forum-Shopping, 80 CORNELL L. REV. 1507 (1995).
[2.] 28 U.S.C. § 1404(a) (1994).
[3.] See Clermont & Eisenberg, supra note 1, at 1511-25.
[4.] See id. at 1515-16.
[5.] See id. at 1525-30.
[6.] Id. at 1530.
[7.] David E. Steinberg, Simplifying the Choice of Forum: A Response to Professor Clermont and Professor Eisenberg, 75 WASH. U. L.Q. 1479, 1480 (1997).
[8.] Id. at 1481-82.
[9.] See Clermont & Eisenberg, supra note 1, at 1511-12.
[10.] See id. at 1521-22 (explaining that the 32% and 4% rates are based on a subset of judgments, so that the actual figures are 29% and 4%).
[11.] Id. at 1517.
[12.] Id. at 1513.
[13.] See id. at 1522-23. Because the number of terminations exceeds the number of judgments, the effect of this difference in non-judgment dismissals is considerably bigger than the opposite effect of that difference in default judgments.
[14.] See id. at 1524 & n.39. We found for diversity cases a drop upon transfer from 50% to 38%. Kevin M. Clermont & Theodore Eisenberg, Do Case Outcomes Really Reveal Anything About the Legal System? Win Rates and Removal Jurisdiction, 83 CORNELL L. REV. 581, 603 n.67 (1998).
[15.] See Clermont & Eisenberg, supra note 1, at 1514 n.18.
[16.] See Clermont & Eisenberg, supra note 14.
[17.] See id. at 604 n.71 (diversity cases only).
[18.] The shift in inconvenience upon removal is probably less than upon transfer, thus implying a relatively smaller removal effect. However, the defendant unilaterally triggers removal, while the defendant must get the judge to approve transfer; this increased opportunity for forum-shopping by the removing defendant would relatively augment the removal effect. The overall result is that the transfer and removal effects for diversity cases are about equal.
[19.] See, e.g., Kimberly Jade Norwood, Shopping for a Venue: The Need for More Limits on Choice, 50 U. MIAMI L. REV. 267, 310 (1996) (urging, as an ideal, tighter venue laws); Brannon J. Buck, Commentary, Forum Shopping and Venue Transfers in Alabama, 48 ALA. L. REV. 671 (1997) (urging greater use of transfer in the state-court system to curtail the growing problem of forum-shopping).
[20.] Steinberg, supra note 7, at 1484.
[21.] See Clermont & Eisenberg, supra note 1, at 1522.
[22.] See Steinberg, supra note 7, at 1484-85, 1491, 1496.
[23.] Id. at 1481 n.6.
[24.] Clermont & Eisenberg, supra note 1, at 1515.
[26.] See Steinberg, supra note 7, at 1486-92, 1496-98.
[27.] See 15 CHARLES ALAN WRIGHT ET AL., FEDERAL PRACTICE AND PROCEDURE §§ 3847-3854 (2d ed. 1986).
[28.] Clermont & Eisenberg, supra note 1, at 1516.
[29.] See id. at 1529.
[30.] See id. at 1529-30.
[31.] See id. at 1530.
[32.] See id. at 1525-30.
[33.] See Steinberg, supra note 7, at 1492-96.
[34.] See David E. Steinberg, The Motion to Transfer and the Interest of Justice, 66 NOTRE DAME L. REV. 443 (1990).
[35.] See Steinberg, supra note 7, at 1499-505.
[36.] 28 U.S.C. § 1404(a) (1994).
[37.] See Clermont & Eisenberg, supra note 1, at 1530.
[38.] See Steinberg, supra note 7, at 1505-12.
[39.] See Clermont & Eisenberg, supra note 1, at 1510 n.10.
[40.] See Kevin M. Clermont, Restating Territorial Jurisdiction and Venue for State and Federal Courts, 66 CORNELL L. REV. 411, 451 (1981).
[41.] Kevin M. Clermont, Whither Due Process? Here To Stay, in ASSOCIATION OF AM. LAW SCHOOLS, AALS CONFERENCE ON CIVIL PROCEDURE 37, 41 (1995).
[42.] Steinberg, supra note 7, at 1512.
[43.] See Steinberg, supra note 7, at 1496-98, 1509-11. But cf. Michael E. Solimine, The Quiet Revolution in Personal Jurisdiction 2-4, 42-46 (1998) (unpublished manuscript, on file with the Washington University Law Quarterly) (showing that recent trends in the law of territorial authority to adjudicate are much more ambiguous than Professor Steinberg's representation).