shows that plaintiffs win much more often when they get to choose the forum. Forum matters Let us define" win rate"as the fraction of plaintiff wins among all judgments for either plaintiff or defendant. Our removal article shows that plaintiffs,win rate in removed cases is very low, compared to original cases in federal court and to state cases. For example, the win rate in original diversity cases is 71%, but for removed diversity cases only 34% The explanation could be the ready one based on the purpose of removal: the defendants thereby defeat the plaintiffs forum advantage and shift thebiases, inconveniences, court quality, and procedural law(but not substantive law)in the defendants' favor. Alternatively, the explanation might lie not in forum impact but instead incase selection" that causes us to be comparing two incomparable groups of cases. Admittedly, removed cases are somewhat different in kind from nonremoved cases. Perhaps, the group of removed cases concentrates the cases that are hard to win, explaining the drop in win rate; that is, removed cases might simply be a set of weak cases involving (i) out-of-state defendants who have satisfied or settled all but plaintiffs' weakest cases or (ii) plaintiff attorneys who have already demonstrated their incompetence by exposing their clients to removal. So, is the explanation forum impact or is it case selection? The observed effect of a dropping win rate after removal prevails across the range of different substantive types of cases, which argues against the drop being a mere case-selection effect Moreover, to isolate the effect on outcome of removal all by itself, one can use the mathematical technique of regression, which is a statistical tool that helps in choosing the nonremoved cases most comparable in kind to the removed cases under study and thus in neutralizing the case-selection effect. Such detailed analysis indicates that forum impact is at work, along with some case selection. After a regression controlling for many case variables-such as circuit, year, case category, amount demanded, procedural development at termination, method of disposition, and kind of subject-matter jurisdiction--the impact of removal remains sizable and significant. The analysis indicates a residual effect of removal, all by itself, that would reduce 50% odds of a plaintiffs win to about 39% in diversity cases. This 11%o reduction from hypothetically even odds represents the impact of the changed forum on the case-the removal effect. In brief, forum really does affect outcome, with removal taking the defendant to a much more favorable forum We also studied the transfer effect, whereby the win rate drops markedly after transfer of venue. Plaintiffs win rate in federal civil cases drops from 58% in cases in which there is no transfer to 29% in transferred cases For the transfer effect, the loss of a favorable forum seems to be the primary explanation, For easy access to part of this database-gathered by the Administrative office of the United States Courts, assembled by the Federal Judicial Center, and disseminated by the Inter-university Consortium for Political and Social Research-see Theodore Eisenberg Kevin M. Clermont, Judicial Statistical Inquiry Form, at http://teddy.law.cornell.edu:8090/questata.htmdiscussedinTheodoreEisenberg&KevinM.ClermontCourtsin Cyberspace, 46 J LEGAL EDUC. 94(1996); see also Kevin M. Clermont Theodore Eisen berg, Litigation Realities, 88 CORNELL L. REV. 119(2002)(summarizing the range of such empirical research) See Kevin M. Clermont Theodore Eisenberg, Xenophilia in American Courts, 109 HARV. L REV. 1120. 1129-32(1996) hereinafter Xenophilia]. Multivariate regression is a statistical technique that quantifies the influence of each of several factors(independent variables)on the phenomenon being studied (dependent variab le). See generally O. FIN in BRUCe LEVIN, STATISTICS FOR LAWYERS chs 13-14(2d ed. 2001)(applying regression analysis to various legal issues)6 26For easy access to part of this database—gathered by the Administrative Office of the United States Courts, assembled by the Federal Judicial Center, and disseminated by the Inter-university Consortium for Political and Social Research—see Theodore Eisenberg & Kevin M. Clermont, Judicial Statistical Inquiry Form, at http://teddy.law.cornell.edu:8090/questata.htm, discussed in Theodore Eisenberg & Kevin M. Clermont, Courts in Cyberspace, 46 J. LEGA L EDUC. 94 (1996); see also Kevin M. Clermont & Theodore Eisenberg, Litigation Realities, 88 COR NELL L. REV. 119 (2002) (summarizing the range of such empirical research). 27See Kevin M. Clermont & Theodore Eisenberg, Xenophilia in American Courts, 109 HARV. L. REV. 1120, 1129-32 (1996) [hereinafter Xenophilia]. Multivariate regression is a statistical technique that quantifies the influence of each of several factors (independent variables) on the phenomenon being studied (dependent variable). See generally MICHAEL O. FINKELSTEIN & BRUCE LEV IN, STATISTICS FOR LAWYERS chs. 13-14 (2d ed. 2001) (applying regression analysis to various legal issues). shows that plaintiffs win much more often when they get to choose the forum.26 Forum matters. Let us define “win rate” as the fraction of plaintiff wins among all judgments for either plaintiff or defendant. Our removal article shows that plaintiffs’ win rate in removed cases is very low, compared to original cases in federal court and to state cases. For example, the win rate in original diversity cases is 71%, but for removed diversity cases only 34%. The explanation could be the ready one based on the purpose of removal: the defendants thereby defeat the plaintiffs’ forum advantage and shift the biases, inconveniences, court quality, and procedural law (but not substantive law) in the defendants’ favor. Alternatively, the explanation might lie not in forum impact but instead in “case selection” that causes us to be comparing two incomparable groups of cases. Admittedly, removed cases are somewhat different in kind from nonremoved cases. Perhaps, the group of removed cases concentrates the cases that are hard to win, explaining the drop in win rate; that is, removed cases might simply be a set of weak cases involving (i) out-of-state defendants who have satisfied or settled all but plaintiffs’ weakest cases or (ii) plaintiff attorneys who have already demonstrated their incompetence by exposing their clients to removal. So, is the explanation forum impact or is it case selection? The observed effect of a dropping win rate after removal prevails across the range of different substantive types of cases, which argues against the drop being a mere case-selection effect. Moreover, to isolate the effect on outcome of removal all by itself, one can use the mathematical technique of regression, which is a statistical tool that helps in choosing the nonremoved cases most comparable in kind to the removed cases under study and thus in neutralizing the case-selection effect.27 Such detailed analysis indicates that forum impact is at work, along with some case selection. After a regression controlling for many case variables—such as circuit,year,case category, amount demanded, procedural development at termination, method of disposition, and kind of subject-matter jurisdiction—the impact of removal remains sizable and significant. The analysis indicates a residual effect of removal, all by itself, that would reduce 50% odds of a plaintiff’s win to about 39% in diversity cases. This 11% reduction from hypothetically even odds represents the impact of the changed forum on the case—the removal effect. In brief, forum really does affect outcome, with removal taking the defendant to a much more favorable forum. We also studied the transfer effect, whereby the win rate drops markedly after transfer of venue. Plaintiffs’ win rate in federal civil cases drops from 58% in cases in which there is no transfer to 29% in transferred cases. For the transfer effect, the loss of a favorable forum seems to be the primary explanation