Jump to ContentJump to Main Navigation
Constitutional Torts and the War on Terror$

James E. Pfander

Print publication date: 2017

Print ISBN-13: 9780190495282

Published to Oxford Scholarship Online: January 2017

DOI: 10.1093/acprof:oso/9780190495282.001.0001

Show Summary Details
Page of

PRINTED FROM OXFORD SCHOLARSHIP ONLINE (www.oxfordscholarship.com). (c) Copyright Oxford University Press, 2018. All Rights Reserved. Under the terms of the licence agreement, an individual user may print out a PDF of a single chapter of a monograph in OSO for personal use (for details see http://www.oxfordscholarship.com/page/privacy-policy). Subscriber: null; date: 20 August 2018

(p.167) Appendix An Empirical Assessment of Bivens Claims

(p.167) Appendix An Empirical Assessment of Bivens Claims

Source:
Constitutional Torts and the War on Terror
Author(s):

Ross J. Corbett

James E. Pfander

Publisher:
Oxford University Press

Introduction

This appendix sets forth the results of a brief study of the success of Bivens litigation in war-on-terror cases. Using Westlaw, we collected a sample of cases in which Bivens claims were asserted in the context of the war on terror. We then evaluated the success of that litigation, not by reference to the degree to which federal courts clarified the governing rules, but by reference to whether the plaintiffs were successful in obtaining some compensation for the alleged violations of their rights.2

Forty-one cases met the criteria for inclusion in the study as Bivens claims arising from some aspect of the war on terror. Of those, nine cases were coded as successes, eight on the basis of settlements apparent on the record, and one in which a voluntary dismissal likely resulted from settlement. The overall success rate of 22 percent bears some resemblance to the success rates identified by other empirical evaluations of Bivens litigation. Claims by military personnel and military contractors as well as domestic prisoner claims and frivolous claims were excluded from the study.

Although the success rate calculation shows that the federal courts do not invariably turn away Bivens claims with war-on-terror overtones, the study confirms many of the key findings in the book. First, none of the Bivens claims involving the military, intelligence services, or extraordinary rendition yielded a successful outcome. That squares with the conclusion (p.168) in Chapter 3 that Bivens litigants have failed in seeking redress or legal vindication in claims brought against the military officials responsible for their detention at Guantanamo Bay and Abu Ghraib. Second, successful claims typically arise from government conduct within the territory of the United States; claims based on foreign conduct were uniformly unsuccessful. Third, defendants were successful in the cases they appealed 100 percent of the time. The striking success of defendants on appeal supports the finding that federal courts have so far refused to uphold pro-plaintiff decisions at the trial court level.

How then can we account for the success that some plaintiffs achieve? Successful cases arose from domestic events, such as flawed airport screening encounters with Transportation Safety Agency (TSA) officials and other domestic law-enforcement settings. These familiar law-enforcement encounters do not implicate the overseas national security and military contexts that have so often led to the judicial deference recounted in this book. Bivens survives, but claimants in overseas war-on-terror cases confront an especially inhospitable federal judiciary.

Design and Methods

Research was conducted by identifying potential war-on-terrorism Bivens suits through Westlaw, rejecting irrelevant or otherwise unsuitable cases, and then examining the complaints and dispositive orders listed on each case’s docket to determine the outcome of each case and identify relevant case characteristics. The research in this project was designed around the null hypothesis that the success rate of war-on-terrorism Bivens suits did not differ appreciably from the success rate of joined, non-prisoner Bivens suits as reported by Reinert.3 The alternative hypothesis was that the success rate of war-on-terrorism Bivens suits was much lower, verging on zero percent.

When a preliminary look at the data suggested that this null hypothesis would be rejected and that the rate of appeal seemed unusually high, additional fields were added to the data set to enable formal testing of three additional null hypotheses. The second null hypothesis thus stated that the rate of appeal would not differ significantly from published rates of appeal in civil cases in federal courts, which range from 10.9 percent to 14.53 percent.4 A third null hypothesis stated the rate of success in the trial court would be independent of the sort of wrongdoing alleged, the place where the plaintiffs first encountered the effects of a federal agent’s decision (usually, the place of arrest), and the appellate circuit in which (p.169) the case was heard. A fourth null hypothesis was that the rate of success on appeal would be independent of the sort of wrongdoing alleged, the place where the plaintiffs first encountered the effects of a federal agent’s decision, and the appellate circuit in which the case was heard.

Case Selection

Potential cases were identified by conducting a search of opinions available on Westlaw in February/March 2015. The search terms were selected to be overinclusive, including general references to “Bivens” and “terrorism” as well as specific keywords related to the war on terrorism, such as “September 11” or “Guantanamo.”5 An attempt was made to capture all war-on-terrorism Bivens suits rather than just a sample, although the resulting data set was of course treated as a sample. Because the significance of the results in this study depends upon the quality of the data set, this Appendix will describe in some detail the criteria employed to exclude cases triggered by the search terms.6

The search yielded 1486 hits, most of which were not germane. For example, every opinion that applied the heightened pleading standard of Ashcroft v. Iqbal7 to whatever claim was before the court after noting in passing that Iqbal had been a Bivens suit involving the war on terrorism triggered a positive search result, as did cases that referred to an event on September 11 of some year, that involved someone named “Bivens,” or that had docket numbers with a 9 close to an 11. Other cases that were easily excluded from the data set involved parties describing each other’s tortious behavior as “terrorism,” general claims of persecution, immigration status adjustment cases, drug cases, many border patrol cases (where terrorism was mentioned to justify tight border controls but not as relevant to the case at bar), and suits brought against terrorist organizations and state sponsors of terrorism by their victims.

Some cases excluded from the search results were related to the war on terrorism, but not to the sort of Bivens suits this study examined. Many of these cases were not Bivens suits at all, naming private parties but no federal defendants or seeking only prospective relief. The latter included suits to unfreeze assets, suits to challenge certain TSA procedures, and suits to gain access to information. Where the injury for which a plaintiff sought retrospective redress was denial of access to information, including an opportunity to be an embedded journalist, the suit was not included in the sample. Additionally, vague claims of discrimination or verbal taunts of “terrorist,” claims by inmates that they were (p.170) subject to special administrative measures because they were classified as “terrorists” where there was no link to terrorism in the facts recited, and suits where the alleged constitutional violation involved suppressing the “truth” about 9/11 were not coded. Also excluded were cases brought by 9/11 cleanup workers for their injuries.

Other categories of cases called for finer judgment as to what could fairly be regarded as a case involving the war on terrorism. Suits brought by protesters objecting to their arrests were excluded from analysis; evaluating the impact of the war on terrorism would require comparing them with protesters in general, an altogether different research project. Cases involving unlawful detention by the TSA were included in the data set, but not those where the injury alleged involved the heightened screening procedures to which all passengers are exposed or routine (if selective) baggage screening. However, if the only ground for the allegedly unlawful detention was that the plaintiff became belligerent upon being subjected to routine baggage screening and similar TSA procedures, the case was excluded—not only are such cases distinguishable from cases brought by people caught up in the war on terrorism, but the search terms were not designed to capture all such cases, and so the inclusion of cases turned up by the search terms would have biased the sample. Suits brought by federal employees concerning security clearances and whistleblower protections were excluded for a similar reason. Suits coded by the Administrative Office of the U.S. Courts as prisoner petitions were excluded because of the uncertain rate at which they are reported in published opinions, as discussed below.8

The exclusion of other categories of cases may prove more controversial. These categories are (1) suits dismissed as frivolous or for failure to prosecute, (2) suits relating to the NSA’s warrantless wiretapping program, and (3) suits brought by soldiers or by civilian defense contractors. Including the first category of cases could bias the data. The search terms yield only those cases that resulted in opinions on motions to dismiss or for summary judgment. One would not expect cases dismissed as frivolous or for failure to prosecute after motions have been filed9 to resemble those dismissed before motions are filed, however. Moreover, in order to obtain success rates in Bivens suits that did not depend to a statistically significant extent upon the makeup of the potential plaintiff population (e.g., upon whether a district encompassed a federal prison), Reinert looked to cases that were not dismissed as frivolous and that had drawn a motion to dismiss or answer.10 Thus, although the exclusions of cases dismissed as frivolous or for want of prosecution will increase success rates, the results will still compare well with previously reported success rates, whereas the (p.171) inclusion of those cases could affect the validity of the data sample. The exclusion of these cases may have an effect on appeal rates, as discussed below.

The second category of cases—those challenging the warrantless wiretapping program attributed to the NSA—was excluded because the legal issues that program raises are in practice distinct from those in other Bivens suits, even when the plaintiff seeks damages under Bivens rather than simply injunctive or declaratory relief. Suits challenging the NSA program tend to focus on the problem of standing. The suits included in this study’s sample, by contrast, arise from a concrete particularized injury to the plaintiffs, and tend to focus on the questions peculiar to Bivens suits: Did the defendants’ conduct violate a constitutional right, are they entitled to qualified immunity, and do special factors counsel hesitation in recognizing a Bivens remedy? Because suits challenging the NSA program encounter different legal hurdles than those applicable to a Bivens suit, the NSA cases were treated as separate populations. Although a broader data set that included these cases could yield interesting results, the exclusion of suits challenging the NSA program should not affect the validity of this Appendix’s findings.

The decision to exclude claims brought by military personnel and the employees of private security companies was a closer call, and future research may fruitfully compare the results of this study with a broader data set including these cases. To be sure, such suits differ from the Bivens claims included in the present study. Members of the armed forces cannot bring Bivens suits against their superior officers,11 and suits brought by civilian security personnel have been blocked on the analogous ground that they would disrupt the military chain of command.12 Still, the results of this study indicate a significant difference between Bivens suits that do not involve the military or intelligence services and those that do, as well as between suits arising on U.S. territory and those arising abroad. In order to test the hypothesis the Bivens special factor of “case[s]‌ involving the military, national security, or intelligence”13 is a bar only when the military or intelligence services are involved, one would have to examine cases that involve military and civilian security personnel (or at least cases where the plaintiff had been accused of collaborating with the enemy, rather than suits where the plaintiff was characterizing service-related injuries or promotion denials as constitutional violations).

Like all studies based on Westlaw-search-and-cull procedures, this one must accept the limitations of the data set: many opinions are not published, and many cases will be missed if only those in the Westlaw database of opinions were included.14 David Hoffman, Alan Ezenman, and (p.172) Jeffrey Lidicker report that only 18 percent of the final orders in their sample produced an opinion of some kind.15 This presents some problems for empirical research based upon published opinions.16 It is for this reason that Reinert identified his cases through PACER rather than through a Westlaw or LEXIS search.17

It is certainly true that the Westlaw-search-and-cull procedure used in this study missed some cases. Cross-checking the list of cases included in this study with a set of appellate and Supreme Court war-on-terrorism Bivens cases, we identified two suits that generated available opinions at the appellate level but not from the district court. In one of these, the case appears not to have generated written opinions at all, the content of the orders appearing as terse docket entries.18 A second case generated written opinions available on PACER, but those opinions are not reflected in the searchable Westlaw database.19 A search of Westlaw appellate cases for this study’s search terms confirms that only these two cases produced appellate opinions without appearing at the district court level. We thus gained some confidence in our case identification protocols, assuming comprehensive Westlaw coverage of appellate court decisions for the period following September 11, 2001. We found a third omitted case, one mentioned in another case that the search criteria did disclose. This case was omitted because it did not generate any opinions prior to being settled.20 These three cases were not included in the study’s sample because they were not discovered using the stated selection criteria, and expanding those criteria to encompass similar cases would undercut the randomness in those criteria.

One should not make too much of the shortcomings of a search-and-cull procedure, however. First, the underreporting is likely not as great as the numbers above suggest. Hoffman et al. drew their data from 2003 and predicted that more opinions would become available online over time.21 Second, it seems reasonable to believe—contrary to the finding by Hoffman et al., that the importance of a case does not increase the odds that an opinion will be published online in a statistically significant way22—that non-frivolous suits alleging constitutional violations in the war on terrorism are in fact special and so are more likely to result in published opinions.23 The case identified above that did not draw a motion to dismiss is unusual enough to justify assuming that not many would be like it.24 Third, of the cases identified through the method used in this study, twenty-six resulted in at least one appeal that was seen to its conclusion, and only the two cases noted above produced appellate opinions but were missed by this study’s case identification method. That is a coverage rate of 92.9 percent, at least of cases that generate appellate opinions. This is (p.173) much higher than would be expected if war-on-terrorism Bivens suits suffered underreporting to the same degree as predicted by Hoffman et al.25 Finally, to the extent that opinions are underreported, Reinert reports that a focus on published opinions results in a lower success rate in Bivens suits.26 Consequently, an undercounting of war-on-terrorism Bivens cases could affect a finding that they are significantly less likely to succeed than Bivens suits in general, not that they are more likely or not significantly different.

The predicted rate at which final orders in prisoner petition cases are reported is astonishingly low, however: 2.6 percent, all other things being equal.27 Even if reporting rates have improved dramatically over the past dozen years, they are likely still too low to permit quantitative research except by identifying sample cases solely by searching dockets. Consequently, prisoner petition cases were excluded from the sample, even when the complaint did not relate to conditions of incarceration.28

We must finally explain how we excluded other cases from the study. Not all persons associated with terrorism have the same goals, and clearly not every innocent person who complains of a constitutional violation in the conduct of an investigation shares salient characteristics with those engaged in terrorism. Thus some sort of distinction between the “war on terrorism” and “domestic” or “homegrown” terrorist plots would introduce a great deal of subjectivity into the sample selection process.29 Nor is it clear that such a distinction would be justified: if we are concerned about judicially enforced constitutional limits on executive action, then the suite of tools employed should determine case selection criteria rather than the characteristics of the plaintiff. Similarly, where the plaintiff alleges involvement by agents associated with the war on terrorism, we include the case even if the defense answers that the actions were unrelated.30 Doing otherwise would require a judgment on the merits of a claim. Last, in order to facilitate comparisons to Reinert’s data for Bivens suits more generally cases were not excluded where all claims against Bivens defendants were dismissed for lack of personal jurisdiction, defective service, expired statute of limitations, and similar reasons.31

Coding Scheme

Coding was done by case, in conformity with other studies that examine success and appeal rates.32 Suits involving multiple plaintiffs are treated as one case and not weighted.33 Where a single group of plaintiffs brought multiple actions that were later consolidated, all the actions are (p.174) treated as one case.34 Similarly, where an action was severed after filing, both subsequent actions are treated as a single case.35 Where a plaintiff brought different suits against different defendants, however, the cases are counted separately.36 Cases brought by separate plaintiffs but later consolidated for pretrial purposes are also treated as separate cases.37

A single person handled case selection, case coding, and accuracy checking.38 Relevant to the present study,39 codes were assigned for the district, the appellate circuit in which the district was located, the disposition of Bivens claims, the type of allegation, the location of the plaintiff’s first interaction with federal authority, the issue of whether the case generated one or more appeals, and the identity of the appellant and result of each appeal. To generate subsets of the data for comparison to other published studies, cases were also assigned codes based on whether a motion to dismiss or answer had been filed, and whether appeals were complete.

The disposition of a case was considered “successful” if it resulted in a settlement agreement or a voluntary dismissal involving any defendant against whom there was still a viable Bivens claim. This conforms to accepted practice regarding studies of success rates, and although in principle counting voluntary dismissals as successes can lead to an overly rosy picture of success rates, that was not a problem for this particular study.40 Because a judgment would be entered in favor of the plaintiff only if the government persisted in taking a losing case to trial or its agents’ actions were so egregious as to warrant summary judgment, there is no reason not to count settlements extracted via litigation as plaintiff successes. The disposition of a case was coded if all Bivens claims had been addressed, even if the case was still open regarding other claims.41 Where only some Bivens claims had been disposed of, however, no disposition was recorded and the case was not included in the analysis.42 In almost every case, the disposition could be determined by looking at the docket or relevant dispositive order.43 Where an order was overturned on appeal, the mandate of the appellate court or subsequent district court order was coded as the case’s disposition.

An allegation was considered to involve the military or intelligence services if it complained of wrongful detention by military authorities, wrongful death as a result of military action, or extraordinary rendition. Extraordinary rendition decisions were assumed (without confirmation) to have been made by the organs of foreign policy rather than domestic law enforcement. For ease of expression, this Appendix speaks of military and civilian allegations, even though some extraordinary rendition claims involve civilian law enforcement.

(p.175) The locus of the complaint was coded into three mutually exclusive categories based on the applicability of the Constitution’s protections, and was usually determined by the place of arrest or search. Codes were assigned for complaints arising on U.S. territory where full constitutional protections apply, complaints arising outside of U.S. jurisdiction or effective control, and complaints arising in a customs and border enforcement context. The last was interpreted expansively to cover stops by Customs and Border Patrol (CBP) agents within U.S. territory,44 but not to cover stops by non-CBP personnel at airports where travel began domestically.45 Cases complaining that the plaintiffs were mistreated abroad at the direction of persons in the United States were treated as raising issues of the extraterritorial application of the Constitution and coded as foreign.

Each Bivens-related appeal in a case was coded by appellant (plaintiff, defendant, or cross-appeal) and result. The coding for results comprised (1) non-merits dismissal; (2) appeal still open; (3) affirmed or dismissed on merits; (4) reversed, remanded, or vacated, in whole or in part; and (5) defendant clearly won cross-appeal. A sixth potential code for a plaintiff’s clear win on cross-appeal was not used. Reversals in full and in part were counted together in order to facilitate comparison to published reversal rates.46 An appellant was considered successful on appeal if it obtained a reversal or reversal in part, or if it was the clear winner in a cross-appeal.

District codes were assigned based on where the case was handled, not on where it was filed (if there was a change of venue). Coding for whether appeals had been exhausted took into account only appeals concerning the Bivens claims.47 Appellate review was considered to have been completed if no appeals were taken and the deadline for filing a notice of appeal had passed, or if the Court of Appeals had issued an opinion in or dismissed the case; requests for rehearing en banc and undecided petitions for certiorari did not prevent a case from being coded as appeal complete.

Data Analysis Methods

The coded data were divided into five overlapping sets: all cases, cases with dispositions, cases with dispositions that had drawn a motion to dismiss or answer (joined cases), cases in which a final judgment had been entered (closed cases), and cases in which appeals were completed. Parallel tests were run on each set, although comparisons to Reinert’s data on Bivens suits in general were run only on joined cases, closed cases, and cases in which appeals had been completed. The set of all cases was relied upon solely to validate the data and for assessing the sample selection process.

(p.176) Most of the tests related to success rates were conducted on the subset of cases with dispositions that had drawn a motion to dismiss or answer, what we follow Reinert in calling “joined cases,”48 even if the viability of other claims prevented the entry of a final judgment. We required cases to have been joined to maintain comparability with his results. And although there are strong theoretical reasons for wanting the outcome of a case to be completely settled before drawing any conclusions from it, such a procedure would have decimated the sample given the length of time it can take to exhaust all appeals. And if one does not insist that the U.S. Supreme Court deny certiorari or issue an opinion in a case, one need not insist that all intermediate-level appeals have been completed in a case before examining it. A similar logic applies to cases where the Bivens claims are no longer before the district court, but the case as a whole is not yet closed.49 Statements about the rate of appeals, on the other hand, are drawn from analysis of all the cases in which a final judgment was entered. It does not make sense to include a disposition not yet ripe for appeal in the denominator of the appeal rate.

Success in a case was cross-tabulated with the appellate circuit, with whether the allegation type involved the military, and with the locus of the alleged tort. Whether an appeal was taken was cross-tabulated with whether the allegation type involved the military. Similarly, reversal rates on appeal were cross-tabulated with appellant, locus of the alleged tort, and whether the allegation type involved the military. Success on appeal was cross-tabulated with appellate circuit. The effect of allegation types, loci, and circuit was measured by partitioning the resulting contingency tables.50 Partial tables were generated for plaintiff success by appellate circuit, given the presence of a military allegation and a foreign locus; appellate success by appellant, given the presence of a military allegation; appellate success by locus, given appellant; and appellate success by circuit, given the presence of a military allegation. Other partial tables were generated, but produced cell counts so low as to provide no meaningful data.

Cross-tabulation was also used to compare success rates with those reported by Reinert for non-prisoner Bivens suits that were not dismissed as frivolous and that drew a motion to dismiss or an answer (35.6 percent).51 Cross-tabulation was chosen over published confidence intervals so that sample size could be taken into account in assaying any divergence. The number of successes was calculated directly from the data Reinert provided in an appendix.52

Cross-tabulations were tested for significance using Barnard’s CSM to perform two-sided exact tests at the 0.05 level.53 Even when the hypothesis (p.177) suggested that a one-sided test might be appropriate, a two-sided test was performed so as not to foreclose the possibility of a surprising result.

Appeal rates were compared with those reported in other studies with binomial probability tests at the 0.05 level with Clopper-Pearson intervals.54 They were compared with the appeal rate for all cases reported by Eisenberg (10.9 percent),55 that for cases with some nonministerial order reported by Hoffman et al. (14 percent),56 and that from all judgments reported by Clermont & Eisenberg (14.53 percent).57 Confidence intervals for success and appellate rates were calculated at the 95 percent level using Agresti & Coull’s “adjusted Wald” interval.58 Analysis was done using R.59

Descriptive Statistics

Fifty-eight cases were identified and coded into the data set, forty-nine of which had all of their Bivens claims resolved and so could be assigned disposition codes. Eight of the cases assigned disposition codes had to be excluded as prisoner petitions. All of the remaining forty-one cases had been joined; one of them still had other claims in the case pending and so was not included in analysis of the appeal rate. These cases were distributed among nineteen judicial districts situated in nine appellate circuits, with about 70 percent of the cases arising within the Second, Ninth, and D.C. Circuits.60 Cases were filed between 2002 and 2013. There was no overlap with the suits that formed the basis of Reinert’s data, which were drawn from cases filed in five districts in 2001–2003.61

Of the cases that were joined, fifteen (36.6 percent) involved allegations concerning the military or intelligence services, as did thirteen (36.1 percent) of the thirty-six appeals taken. Most of the cases involved allegations of wrongdoing in the United States (twenty-two, or 53.7 percent), as did most of the appeals (twenty-one, or 58.3 percent). More than a quarter arose on foreign territory (twelve cases (29.3 percent), ten appeals (27.8 percent)), with the rest arising in a customs/border enforcement context (seven cases (17.1 percent), five appeals (13.9 percent)). Only one case went to trial, and that resulted in a judgment for the defendants. Settlement was apparent on the record for eight of the forty-one cases that had been joined. One case was voluntarily dismissed under circumstances making it probable that some sort of settlement was reached.62 Twenty-three (63.9 percent) of the appeals were taken by the plaintiff, eleven (30.6 percent) by the defendant, and two (5.6 percent) were cross-appeals. (p.178) Of these, eight appeals were dismissed, four were still open when the data were analyzed, fourteen affirmed the decision below, eight reversed it in whole or in part, and both cross-appeals were wins for the defendants.

Plaintiffs’ Success in Obtaining Redress

The success rate of Bivens suits involving the war on terrorism was 22.0 percent. (See Table 1) This success rate is outside the confidence interval for joined non-frivolous cases reported by Reinert.63 However, cross-tabulation of war-on-terrorism Bivens suits with Reinert’s data does not permit one to conclude that the success rate of the former differs significantly from that of the latter.64

Moreover, within war-on-terrorism Bivens suits, claims involving the military and intelligence services significantly differ from other claims.65 The success rate of the latter is 34.6 percent, that of the former, 0.0 percent. When all cases are examined, the success rate of cases heard within the Second Circuit is higher than in the rest of the circuits, that of those heard in the D.C. Circuit, lower, but the different rates of success cease to be significant when military allegations are excluded from analysis.66

Table 1. Plaintiff Success in War-On-Terrorism Bivens Cases

Cases

Successes

Rate

95% CI

All Joined Cases

41

9

22.0%

11.8–36.1

By Allegation:

Nonmilitary

26

9

34.6%

19.3–53.9

Military

15

0

0.0%

<0.0–23.9

By Locus:

Non-foreign

29

9

31.0%

17.1–49.4

Foreign

12

0

0.0%

<0.0–28.2

A similar difference in success rate distinguishes claims arising in foreign territory from those arising on U.S. territory, which includes areas governed by the rules regarding customs and border enforcement.67 Foreign locus claims overlap with those alleging some wrongdoing by the military and intelligence services in all but three cases.68 No plaintiff complaining of actions abroad has succeeded, whereas other plaintiffs obtain a settlement or dismiss their cases voluntarily 31 percent of the time. (p.179)

Thus, excluding prisoner petitions, the success rate of war-on-terrorism Bivens cases that do not involve the military or intelligence services (34.6 percent) is almost indistinguishable from that observed regarding Bivens cases that survive sua sponte judicial screening (35.6 percent).69 For cases that involve the military or intelligence services, on the other hand, no successes were identified.70

Table 2. Appeal Rates of War-On-Terrorism Bivens Cases

Cases

Appeals

Rate

95% CI

All Allegations

41

32

80.0%

65.0–89.8

Nonmilitary

26

19

76.0%

56.2–88.8

Military

15

13

86.7%

60.9–97.5

Involvement of the Appellate Courts

The appeal rate in all war-on-terrorism Bivens cases is 80.0 percent. That for cases without a military allegation, 76.0 percent. That for cases with military allegations, 86.7 percent. (See Table 2). Although the rate at which cases involving the military are appealed does not differ significantly from that concerning allegations against civilians,71 all three rates are much, much higher than the rate reported by Eisenberg for all cases (10.9 percent),72 that reported by Hoffman et al. for cases with some nonministerial order (14 percent),73 and that reported by Clermont & Eisenberg for all cases with judgments (14.53 percent).74

For Bivens suits arising from the war on terrorism, appeal rates and reversal rates are significantly higher than those for other civil rights cases heard in the federal courts.75 (See Table 3). Appellants in this study obtained a reversal or reversal in part in 36.4 percent of the appeals that were seen through to a decision, as opposed to 13.5 percent in civil rights cases more generally.76

There was a sharp distinction in the reversal rate when looking to allegation type, locus, and which party filed an appeal, however.77 Where a military allegation was involved, the reversal rate dropped to 10.0 percent, largely indistinguishable from the reversal rate for civil rights cases in general.78 District court orders in cases not involving military allegations, on the other hand, were reversed 58.3 percent of the time.79 A similar division appears between cases arising on foreign territory on the one (p.180) hand and those arising in the United States or in a border patrol context on the other, with the latter being reversed 53.3 percent of the time, the former, none.81 In every case, moreover, regardless of allegation type or locus, defendants obtained reversals far more frequently, as is the case in both civil cases and civil rights cases more generally.82 Although the poor success rate of plaintiffs on appeal in war-on-terrorism Bivens cases roughly corresponds to their success rate more generally,83 the 100 percent success rate of defendants on appeal more than doubles their advantage in ordinary civil rights cases.84

Table 3. Reversal Rates of Appeals Decided on Merits

Appeals80

Reversals

Rate

95% CI

All Appeals

22

8

36.4%

19.6–57.1

military

10

1

10.0%

0.4–42.6

nonmilitary

12

7

58.3%

31.9–80.7

foreign

7

0

0.0%

5.0–40.4

non-foreign

15

8

53.3%

30.1–75.2

Plaintiff Appellant

18

2

11.1%

1.9–34.0

Defendant Appellant

8

8

100.0%

62.8–100+

Notes:

(1.) Ross J. Corbett, J.D. 2015, Northwestern University School of Law, PhD 2005, University of Toronto. Thanks to Josh Fischman for help with the statistics and to Alexander Reinert for discussions of the methods of case selection and coding he used in a prior study.

(2.) Full tables and extended analysis are omitted. Instead, relevant p-values are reported in the notes. Except where a particular docket entry is referred to, cases are given by docket number and date of final disposition rather than by the page and volume of an opinion in the Federal Supplement, even where the latter is available. Tables 1 and 2 at the end of these endnotes set forth the relevant docket information.

(3.) See Alexander A. Reinert, Measuring the Success of Bivens Litigation and Its Consequences for the Individual Liability Model, 62 Stan. L. Rev. 809, 841 n.154 (2010) (rate of 21/59, or 35.6 percent).

(4.) See Kevin M. Clermont & Theodore Eisenberg, Plaintiphobia in the Appellate Courts: Civil Rights Really Do Differ from Negotiable Instruments, 2002 U. Ill. L. Rev. 947, 967 tbl.5 (2002) (14.53 percent of all cases with judgments); David Hoffman et al., Docketology, District Courts, and Doctrine, 85 Wash. U. L. Rev. 681, 711 (2007) (14 percent of all cases with some nonministerial order); Theodore Eisenberg, Appeal Rates and Outcomes in Tried and Nontried Cases: Further Exploration of Anti-Plaintiff Appellate Outcomes, 1 J. Empirical Legal Stud. 359, 664 tbl.1 (2004) (10.9 percent for all cases). We were unable to find published appeal rates for Bivens cases or for a broader category of all constitutional tort cases. Consequently, the appeal rate could be compared only with civil cases in general.

(5.) The selection of search terms necessarily involved some guesswork as to which words and phrases would appear in relevant opinions, and as a result the search expanded several times after cases had been identified as more comprehensive search terms came to light. Search terms were not removed, even if new terms seemed more comprehensive, in order to ensure that the cases selected all met the stated criteria for data selection. The ultimate search string parses to ((“Abu Ghraib” Iraq Afghanistan Guantanamo “September 11” (9 /1 11) terroris!) & Bivens) (“no-fly list” “Terrorist Screening Center”) ((“Abu Ghraib” Iraq Afghanistan Guantanamo) & (terroris! (war /s terror!))) ((rendition interrogat!) /p (terroris! (war /s terror!))) & ((summary /s judgment) (motion /s dismiss)), with results further narrowed to documents from federal trial courts, civil cases only, and dated after 9/11/2001. The inclusion of terms relating to summary judgment and motions to dismiss was designed to reduce the number of opinions any one case would generate.

(6.) A separate search was not conducted via LEXIS, as there is no easy way to export the results of both searches and compare the results to detect any differences. But see Hoffman et al., supra note 4, at 710 & n.138 (reporting results of comparison (p.223) of Westlaw and LEXIS coverage, noting that all cases in the sample reported in LEXIS were also reported in Westlaw, but not vice versa).

(7.) Ashcroft v. Iqbal, 556 U.S. 662 (2009).

(8.) Because the decision to exclude these cases was prompted by Hoffman et al., supra note 4, the same definition of a prisoner petition was used, namely, one coded as 510, 530, 535, 540, 550, or 555. See id. at 749.

(9.) E.g., Iqbal v. FBI, No. 11-CV-00369 (M.D. Fla. June 5, 2014) (dismissed sua sponte following plaintiff’s failure to respond to motion for summary judgment; plaintiff was not Javad Iqbal).

(10.) See Reinert, supra note 3, at 841–43, 862 app. tbl. 9.

(11.) United States v. Stanley, 483 U.S. 669, 679–84 (1987); Chappell v. Wallace, 462 U.S. 296, 304 (1983).

(12.) E.g., Vance v. Rumsfeld, 701 F.3d 193, 198–203 (7th Cir. 2012); Doe v. Rumsfeld, 683 F.3d 390, 395–96 (D.C. Cir. 2012).

(13.) Doe, 683 F.3d at 394.

(14.) See Hillel Y. Levin, Making the Law: Unpublication in the District Courts, 53 Vill. L. Rev. 973, 985 (2008).

(15.) Hoffman et al., supra note 4, at 710 n.139. Levin reports that only 3 percent of orders result in a written opinion, but that number reflects every order entered in a case, including scheduling orders, and seems to be taken from Hoffman et al. See Levin, supra note 14, at 985.

(16.) See Hoffman et al., supra note 4, at 727–29.

(17.) Cf. Reinert, supra note 3, at 831–32 (noting deficiency in past research based on analysis of published cases).

(18.) See Unus v. Kane, No. 04-CV-00312 (E.D. Va. Nov. 2, 2007).

(19.) See Mirmehdi v. Ashcroft, No. 06-CV-05055 (C.D. Cal. May 26, 2009).

(20.) See Habeeb v. United States, No. 06-CV-00454 (W.D. Wash. July 3, 2007). This case was referred to in a joint motion to vacate judgment for the defense in another case filed by the same plaintiff over the same events. Joint Motion to Vacate at 3, Habeeb v. Castloo, (D. Mont. Jul. 12, 2007) (No. 05-CV-00024).

(21.) Hoffman et al., supra note 4, at 708, 729.

(22.) Id. at 719.

(23.) Testing this assumption would likely be impracticable, unless the rate at which opinions are published in war-on-terrorism cases is abnormally low. In principle, one could replicate Reinert’s procedure for identifying Bivens claims for a subset of districts and years, checking that subset for any cases that were not otherwise identified. Over the time period of 9/11/01–03/05/15, however, only fifty-eight cases met the criteria to be coded (and many of these had to be excluded from the data sample). These were spread across twenty-three U.S. District Courts, a mean of 2.5 cases per district, with the median and mode both being 1. The District Court for the District of Columbia had the highest number of cases at eleven, but even that is an average of less than one per year. Even if the rate at which cases went unreported has remained unchanged since 2003 (viz. 18 percent), that would still yield only 322 codeable cases spread over fourteen years and filed in an unknown number of district courts. In practice, then, finding that every war-on-terrorism Bivens case in a sample district/year was turned up in this study’s search would not tell us much unless the sample size was so large as to be unwieldy.

(24.) There was no opinion in the case—and thus no way of detecting it by searching for opinions—because it was settled without drawing a defense motion to (p.224) dismiss. The defendants did answer, however, so the case would have counted as joined and so have been included in this sample, had it been discovered in a regular fashion. See Habeeb v. United States, No. 06-CV-00454 (W.D. Wash. July 3, 2007). But this case was special. It was filed a year after another Bivens suit was filed by the same plaintiff against different defendants relating to the same incident. The government fought that earlier case, obtained summary judgment, and then (while that judgment was being appealed) conceded error and asked the court to vacate the judgment in its favor. Joint Motion to Vacate, Habeeb v. Castloo, (D. Mont. July 12, 2007) (No. 05-CV-00024). That case, because it did generate a motion for summary judgment, was detected by the search conducted for this research. The missed case is thus the exception that proves the rule that, normally, if the government does not concede error before a war-on-terrorism Bivens case is filed, it will at least file a motion to dismiss before conceding error.

(25.) The magnitude of a 92.9 percent appellate opinion coverage rate should not be overstated, however. The unit of analysis adopted by Hoffman et al., supra note 4, was the disposition, not the case, so a predicted opinion-writing rate of 23.6 percent for final dispositions in baseline jurisdictions and in cases not involving private torts, contract, prisoner petitions, or labor & employment disputes is not a prediction of whether any opinion at all will be written in that case. Cases with multiple dispositions are correspondingly more likely to have an opinion associated with them. Moreover, Hoffman et al., supra, coded for whether an appeal was taken in a case, whereas only about half of all appeals are pursued to a judgment on the merits. See Eisenberg, supra note 4, at 660, 664 tbl.1. On the other hand, dispositions that may or may not be immediately appealable (e.g., rulings on motions to dismiss) generate opinions at about half the rate of final dispositions. Hoffman et al., supra, at 746 tbl.B-4. Determining the probability that a type of case (rather than a disposition in that type of case) would produce an opinion would require calculating an anticipated number of intermediate, final, and management orders for that type of case. Even then, the dismissal rate for appeals of that kind of case would be an estimate. Nevertheless, as a subjective judgment, capturing 92.9 percent of cases that produced an appellate opinion by searching for district court opinions seems high in comparison with the opinion-writing rates predicted by Hoffman et al.

(26.) Reinert, supra note 3, at 813–14.

(27.) Hoffman et al., supra note 4, at 746 tbl.B-4.

(28.) These cases were coded, but they were not included in any data reported in this Appendix.

(29.) See, e.g., Hughes v. McWilliams, No. 04-CV-07030 (S.D.N.Y. Dec. 16, 2009) (arrested after threatening email led to concerns of a terrorist plot to poison water supply); Lonegan v. Hasty, No. 04-CV-02743 (E.D.N.Y. Oct 8, 2009) (monitoring of attorney-client communication of those held in war on terrorism); Hatfill v. Ashcroft, No. 03-CV-01793 (D.D.C. Mar. 16, 2007) (dismissing Bivens claims for conduct during investigation into 2001 anthrax attacks). A distinction based on characteristics of the plaintiffs would conceal cases where the plaintiffs allege that the unconstitutional profiling was also misdirected. E.g., Kurzberg v. Ashcroft, No. 04-CV-03950 (E.D.N.Y. Sept. 25, 2006) (Israelis allegedly mistaken for Arabs and detained after 9/11).

(30.) E.g., Abdullah v. Arena, No. 12-CV-14766 (E.D. Mich. Apr. 3, 2014) (plaintiff shot in weapons sting operation involving agents of FBI Detroit Joint Terrorism Task (p.225) Force); cf. Bhutta v. Bush, No. 05-CV-70433 (E.D. Mich. Mar. 7, 2006) (plaintiff alleged arrest was part of war-on-terrorism investigation; defendants answered that investigation had concluded and arrest was for separate crime; not included in sample because coded as prisoner petition).

(31.) E.g., Abdullah v. Arena, No. 12-CV-14766 (E.D. Mich. Apr. 3, 2014); Kurzberg v. Ashcroft, No. 04-CV-03950 (E.D.N.Y. Sept. 25, 2006); Islamic Relief Agency v. Ashcroft, No. 04-CV-02264 (D.D.C. Sept. 15, 2005); see also Hamad v. Gates, No. 10-CV-00591 (W.D. Wash. Nov. 2, 2010) (granting motion to dismiss for lack of personal jurisdiction for only some Bivens defendants).

(32.) See, e.g., Reinert, supra note 3, at 832–34; Clermont & Eisenberg, supra note 4, at 950–51; Eisenberg, supra note 4, at 661–63; see also Hoffman et al., supra note 4, at 711 (reporting appeal rate for cases rather than for dispositive orders).

(33.) In principal, one co-plaintiff might settle while the other goes on to lose the case. This problem did not arise in this study, although it should be noted that the coding scheme described here does not account for that possibility. Cf. Elmaghraby v. Ashcroft, No. 04-CV-01809 (E.D.N.Y May 3, 2004) (settlement by Elmaghraby, leaving co-plaintiff Iqbal to pursue case to Supreme Court before eventually settling remaining Bivens claims).

(34.) E.g., In re Iraq & Afghanistan Detainees Litig., No. 06-MC-00145 (D.D.C. Mar. 27, 2007), aff’d sub nom. Ali v. Rumsfeld, 649 F.3d 762 (D.C. Cir. 2011).

(35.) E.g., Al-Kidd v. United States, No. 05-CV-00093 (D. Idaho Jan. 13, 2015); Al-Kidd v. Sugrue, No. 06-CV-01133 (W.D. Okla. Mar. 28, 2008).

(36.) E.g., Khorrami v. Mueller, No. 07-CV-00812 (E.D. Wis. Dec. 15, 2014) (judgment for defendants); Khorrami v. Rolince, No. 03-CV-06579 (N.D. Ill.) (still pending when data was analyzed).

(37.) E.g., Allaithi v. Rumsfeld, No. 08-CV-01677 (D.D.C. Feb. 1, 2013); Celikgogus v. Rumsfeld, No. 06-CV-01996 (D.D.C. Feb. 1, 2013).

(38.) Over the course of research the data set migrated from Excel to SPSS to a comma-delineated file for use in R. Migration was validated by spot-checking and an analysis of the resulting descriptive statistics.

(39.) The initial coding scheme was simplified to the one described in the body of the text by combining table columns. Although Graeme D. Ruxton & Markus Neuhäuser, Good Practice in Testing for an Association in Contingency Tables, 64 Behavioral Ecology & Sociobiology 1505, 1509 (2010), caution against combining table columns with low cell counts when there were reasons to count the columns separately in designing the research, several of the distinctions that defined the original coding scheme were unnecessarily fine. Moreover, some non-results also turned out to be uninteresting with respect to this study’s primary focus, and the low number of cases involved means that a lack of a significant finding is not particularly probative.

(40.) See Reinert, supra note 3, at 833 (counting settlement and voluntary dismissal as success); Stewart Schwab & Theodore Eisenberg, Explaining Constitutional Tort Litigation: The Influence of the Attorney Fees Statute and the Government as Defendant, 73 Cornell L. Rev. 719, 726–28 (1988) (same); cf. Theodore Eisenberg & Stewart Schwab, The Reality of Constitutional Tort Litigation, 72 Cornell L. Rev. 641, 683–84 (1987) (noting that they had to reduce their reported success rate after contacting attorneys for parties with uncertain dispositions). In principle, judgments for the plaintiff should also be counted as successful, but none occurred in this data set.

(41.) E.g., Tarhuni v. Holder, No. 13-CV-00001 (D. Or. Mar 26, 2014) (dismissing claims against individual-capacity defendants).

(42.) E.g., Turkmen v. Ashcroft, No. 02-CV-02307 (E.D.N.Y Jan. 15, 2013), appeal docketed, No. 13-1662 (2d Cir. Apr. 25, 2013) (granting motion to dismiss in part and denying in part).

(43.) The exception was Mayfield v. United States, No. 04-CV-01427, at 4 (D. Or. Sep. 26, 2007), in which a court order on a motion for summary judgment concerning the plaintiff’s request for declaratory judgment mentioned in the procedural history that a stipulated settlement had previously been reached regarding the Bivens claims.

(44.) E.g., Habeeb v. Castloo, No. 05-CV-00024 (D. Mont. July 12, 2007) (stopped on a train traveling through Montana).

(45.) E.g., Ibrahim v. Dep’t Homeland Security, No. 06-CV-00545 (N.D. Cal. Mar 15, 2010) (stopped at airport while returning to Malaysia).

(46.) See Clermont & Eisenberg, supra note 4, at 951. (2002)

(47.) For example, Ibrahim v. Dep’t Homeland Security, No. 06-CV-00545 (N.D. Cal. Mar 15, 2010) (settling Bivens claims), was coded as having its appeals complete, even though an appeal was still pending on the question of attorney’s fees.

(48.) This followed Reinert’s coding scheme for a “joined” claim. In contrast, much the literature considers a case joined only when the answer is filed. Like Reinert, we care a great deal about cases that are dismissed before an answer is required. See Reinert, supra note 3, at 841–42 & n.153.

(49.) This point is of little practical effect, however. As noted below, the set of joined cases is larger than the set of closed cases by one.

(50.) See Alan Agresti, Categorical Data Analysis 3.3.3–5 (3d ed. 2013) (discussing partitioning).

(51.) See Reinert, supra note 3, at 841 n.154 (reporting twenty-one successes out of fifty-nine cases not relating to prison conditions).

(52.) Id. at 853–58 tbl. 1. Data was copied by cutting and pasting from the Westlaw digitization of the article. As confirmed with Prof. Reinert by email, the Y and N under “Issued Joined” must be swapped.

(53.) All of the contingency tables in this study are both 2x2 and unconditioned, and most have a cell with an expected value of less than 5. But Pearson’s χ‎2 should not be used for tables where more than 20 percent of cells have an expected value of less than 5 or where the expected value of any cell is less than 1, and Fisher’s exact test might be overly conservative regarding unconditioned 2x2 contingency tables. Where neither is appropriate, Ruxton and Neuhäuser recommend a Fisher’s Exact-Boschloo test or Barnard’s CSM test. The latter is more powerful, but computationally exacting. Ruxton & Neuhäuser, supra note 39, at 1507–09; cf. Devan V. Mehrotra, Ivan S.F. Chan & Roger L. Berger, A Cautionary Note on Exact Unconditional Inference for a Difference between Two Independent Binomial Proportions, 59 Biometrics 441, 447–49 (2003) (drawback of Barnard’s test is computational feasibility). The software package used in this study to calculate Barnard’s CSM performs only a test of homogeneity, not of independence, but the result of this is increased conservatism and so should not affect the validity of the results. Cf. John Ludbrook, Analysis of 2x2 Tables of Frequencies: Matching Test to Experimental Design, 37 Int’l J. Epidemiology 1430, 1432 (2008) (tests designed for comparative trials are too conservative when dealing with a double dichotomy trial, but exact tests are not feasible for double dichotomy trials).

(54.) Clopper-Pearson intervals were chosen because the small sample size would have made Wald intervals too prone to Type I error. See Alan Agresti & Brent Coull, Approximate Is Better than “Exact” for Interval Estimation of Binomial Proportions, 52 Am. Statistician 119 (1998).

(55.) Eisenberg, supra note 4, at 664 tbl. 1.

(56.) Hoffman et al., supra note 4, at 711.

(57.) Clermont & Eisenberg, supra note 4, at 967 tbl. 5.

(58.) See Agresti & Coull, supra note 54, at 122. The Clopper-Pearson intervals used in performing binomial probability tests are not reported because their narrow range would be misleading. Cf. id. at 125; Agresti, supra note 50, at 16.6.1 (Clopper-Pearson good for hypothesis testing but not interval estimation). A series of tests performed on the data in this study confirmed that Agresti-Coull intervals best approximated 0.95 coverage on a consistent basis. By contrast, the actual coverage of the asymptotic (unadjusted Wald) interval for appeals taken in all closed cased was .905; that of the Clopper-Pearson (exact) interval for appeals taken in closed cases involving military allegations was .990. In no case does the Agresti-Coull interval reported in this Appendix have a coverage deviating from 0.95 by more than 0.022, except for cases in which every method produces a coverage >.999.

(59.) R: A language and environment for statistical computing, Version 3.1.3 (last visited May 2015), available at http://www.R-project.org/. Agresti-Coull (“adjusted Wald”) confidence intervals were calculated using the Binomial Confidence Interval package for R written by Sundar Dorai-Raj, Version 1.1-1 (2014), available at http://cran.r-project.org/web/packages/binom/. Note that the Binomial Confidence Interval package does not use the “add two success and two failures” approximation of the Agresti-Coull interval. Fisher-Boschloo and Barnard’s CSM were calculated using the Unconditional Exact Test package for R written by Peter Calhoun, Version 1.5 (2015), available at http://cran.r-project.org/web/packages/Exact/.

(60.) No result was significant in one set of cases (joined, closed, or appeals completed) but not in the others.

(61.) Cf. Reinert, supra note 3, at 832.

(62.) In that case, the court entered judgment for the defendants. The plaintiff then took an appeal. The appeal was dismissed by stipulation of the parties, the government conceded that it had been in the wrong, the judgment for the defendants was vacated by the district court upon a joint motion, and the case was not reopened. See Habeeb v. Castloo, No. 05-CV-00024 (D. Mont. July 16, 2007). A companion case not included in this sample settled around the same time. Habeeb v. United States, No. 06-CV-00454 (W.D. Wash. July 3, 2007).

(63.) Reinert provides a 95 percent confidence interval for all joined non-frivolous cases, but does not break out non-prisoner cases. See Reinert, supra note 3, at 862 App. tbl. 9 (reporting 95 percent CI of 22.5–38.4 percent, which corresponds to the asymptotic (Wald) confidence interval). The corresponding number for just non-prisoner cases is 23.4–47.8 percent; the Agresti-Coull (adjusted Wald) confidence interval would be 24.6–48.4 percent. Cf. Reinert, at 841 n.154 (listing success rate by case type).

(64.) Two-sided Barnard’s CSM p = .152.

(65.) Two-sided Barnard’s CSM p = .007.

(66.) For a cross-tabulation of trial-court success rate for cases heard within the Second Circuit and the success rate for cases heard in other circuits, two-sided (p.228) Barnard’s CSM p = .027 for all cases but .172 for nonmilitary cases. For the corresponding cross-tabulation regarding the D.C. Circuit, p = .036 for all cases but .566 for those not involving the military or intelligence services.

(67.) Two-sided Barnard’s CSM p = .025. The success rate for claims arising in a customs and border enforcement context is roughly that for claims arising on U.S. territory—28.6 percent and 31.8 percent, respectively—with only the distinction between foreign and non-foreign loci being significant.

(68.) Two cases involved Jose Padilla, who was arrested in Chicago but held in military custody. See Padilla v. Yoo, No. 08-CV-00035 (N.D. Cal. July 9, 2012); Lebron v. Rumsfeld, No. 07-CV-00410 (D.S.C. Feb. 18, 2011). A third was filed by Maher Arar, who was stopped by immigration officials in New York, deported, and subsequently tortured in Syria. See Arar v. Ashcroft, No. 04-CV-00249 (E.D.N.Y. July 28, 2006).

(69.) Cf. Reinert, supra note 3, at 841 n.154 (reporting twenty-one successes out of fifty-nine non-prisoner, non-frivolous joined Bivens cases). Two-sided Barnard’s CSM p = .859.

(70.) Two-sided Barnard’s CSM p = .003.

(71.) Two-sided Barnard’s CSM p = .486.

(72.) Two-sided exact binomial probability test with expected appeal rate as reported by Eisenberg, supra note 4, at 664 tbl. 1: p < .000 regardless of allegation type.

(73.) Two-sided exact binomial probability test with expected appeal rate as reported by Hoffman et al., supra note 4, at 711: p < .000 regardless of allegation type.

(74.) Two-sided exact binomial probability test with expected appeal rate as reported by Clermont & Eisenberg, supra note 4, at 967 tbl. 5: p < .000 regardless of allegation type.

(75.) Some cases generated multiple appeals, but in each case only one appeal was seen through to a decision or dismissal on the merits.

(76.) Two-sided exact binomial probability test with expected reversal rate as reported by Clermont & Eisenberg, supra note 4, at 967 tbl. 5: p = .006.

(77.) There was no significant difference among circuits in whether a plaintiff or defendant was successful in obtaining a reversal on appeal, perhaps due to the small number of cases involved.

(78.) Two-sided exact binomial probability test comparing reversal rate of military allegations (1/10) with expected reversal rate in civil rates cases of 13.52 percent reported by Clermont & Eisenberg, supra note 4, at 967 tbl. 5: p > .999.

(79.) Two-sided exact binomial probability test comparing reversal rate of nonmilitary allegations (7/12) with expected reversal rate in civil rights rates cases of 13.52 percent reported by Clermont & Eisenberg, supra note 4, at 967 tbl. 5: p < .000.

(81.) Two-sided Barnard’s CSM p = .016.

(82.) See Clermont & Eisenberg, supra note 4.

(83.) Two-sided exact binomial probability test comparing reversal rate obtained by plaintiff appellants (2/18) with expected reversal rate in civil rights cases of 12.54 percent reported by id. at 967 tbl. 5: p > .999.

(84.) Two-sided exact binomial probability test comparing reversal rate obtained by defendant appellants (8/8) with expected reversal rate in civil rights cases of 46.05 percent reported by Clermont & Eisenberg, supra note 4, at 967 tbl. 5: p = .002. (p.232) (p.231) (p.230) (p.229) (p.233)

(Table 1.) Success Rate Data

No.

Dist

Docket

Disp

Joined

PP

Success

Prose

App_Comp

Allege

App_Taken

Allege_Mil

Locus

1

E.D. Mich.

12-CV-14766

Def Dispositive Motion Granted

Y

N

N

N

Y

CWD

Y

N

US

2

D. Nev.

08-CV-01096

Def Dispositive Motion Granted

Y

N

N

N

Y

WCD

N

N

CB

3

N.D. Ohio

10-CV-01492

Def Dispositive Motion Granted

Y

N

N

N

Y

WCD

Y

N

CB

4

N.D. Ill.

05-CV-03761

Def Dispositive Motion Granted

Y

N

N

N

Y

TSA

Y

N

CB

5

S.D.N.Y.

09-CV-09066

Def Dispositive Motion Granted

Y

N

N

Y

Y

CI

Y

N

US

6

D.D.C.

12-CV-01192

Def Dispositive Motion Granted

Y

N

N

N

Y

WDM

N

Y

FOR

7

D. Or.

13-CV-00899

.

.

.

.

.

.

.

N

.

8

E.D.N.Y.

04-CV-03950

Def Dispositive Motion Granted

Y

N

N

N

Y

WCD

Y

N

US

9

E.D. Mich.

12-CV-11656

.

.

.

.

.

.

.

N

.

10

E.D.N.Y.

05-CV-03919

Settlement

Y

N

Y

N

Y

WCD

Y

N

CB

11

E.D.N.Y.

02-CV-02307

.

.

.

.

.

.

.

N

.

12

E.D. Mich.

07-CV-13678

Def Dispositive Motion Granted

Y

N

N

N

Y

CI

Y

N

US

13

M.D. Fla.

11-CV-00369

.

.

.

.

.

.

.

N

.

14

N.D. Cal.

06-CV-00545

Settlement

Y

N

Y

N

Y

TSA

Y

N

US

15

D. Or.

13-CV-00001

Def Dispositive Motion Granted

Y

N

N

N

N

TSA

N

N

CB

16

W.D. Ok.

06-CV-01133

.

.

.

.

.

.

.

N

.

17

D.D.C.

10-CV-01996

Settlement

Y

Y

Y

N

Y

TI

N

N

18

D. Idaho

05-CV-00093

Settlement

Y

N

Y

N

Y

WCD

Y

N

US

19

E.D. Va.

05-CV-01417

Def Dispositive Motion Granted

Y

N

N

N

Y

ER

Y

Y

FOR

20

C.D. Cal.

11-CV-00301

Def Dispositive Motion Granted

Y

N

N

N

N

CI

Y

N

US

21

E.D. Va.

12-CV-00823

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

22

D.D.C.

06-CV-01996

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

23

D.D.C.

08-CV-01677

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

24

W.D. Wash.

10-CV-00591

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

25

N.D. Cal.

08-CV-00035

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

US

26

D.S.C.

07-CV-00410

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

US

27

S.D.N.Y.

08-CV-03592

Voluntary Dismissal

Y

Y

Y

Y

Y

TI

N

N

28

S.D.N.Y.

04-CV-07030

Def Dispositive Motion Granted

Y

N

N

Y

Y

WCD

Y

N

US

29

D. Or.

04-CV-01427

Settlement

Y

N

Y

N

Y

CI

Y

N

US

30

S.D. Ill.

12-CV-01198

.

.

.

.

.

.

.

N

.

31

S.D. Ill.

13-CV-00456

.

.

.

.

.

.

.

N

.

32

S.D. Ill.

13-CV-00866

.

.

.

.

.

.

.

N

.

33

S.D. Ill.

13-CV-01017

Voluntary Dismissal

N

Y

Y

Y

Y

TI

N

N

34

D. Mont.

05-CV-00024

Voluntary Dismissal

Y

N

Y

N

Y

WCD

Y

N

CB

35

S.D.N.Y.

02-CV-09802

Settlement

Y

N

Y

N

Y

WCD

Y

N

US

36

E.D.N.Y.

09-CV-00126

Def Dispositive Motion Granted

Y

Y

N

Y

Y

TI

N

N

37

E.D. Mich.

13-CV-10757

Def Dispositive Motion Granted

Y

Y

N

Y

Y

CI

N

N

38

S.D.N.Y.

13-CV-02072

Def Dispositive Motion Granted

Y

Y

N

N

N

TI

N

N

39

N.D. Cal.

07-CV-00200

Def Dispositive Motion Granted

Y

N

N

N

Y

WCD

N

N

US

40

D.D.C.

03-CV-01793

Def Dispositive Motion Granted

Y

N

N

N

Y

CI

N

N

US

41

N.D. Ga.

06-CV-01720

Def Dispositive Motion Granted

Y

N

N

N

Y

CI

Y

N

US

42

E.D. Wis.

07-CV-00812

Def Judgment

Y

N

N

N

N

WCD

Y

N

US

43

D. Col.

03-CV-01959

Def Dispositive Motion Granted

Y

Y

N

N

Y

TI

Y

N

45

E.D. Mich.

05-CV-70433

Def Dispositive Motion Granted

Y

N

N

Y

Y

WCD

N

N

US

46

S.D.N.Y.

04-CV-07772

Def Dispositive Motion Granted

Y

Y

N

Y

Y

CI

N

N

47

W.D. La.

02-CV-01933

Settlement

Y

N

Y

N

Y

WCD

Y

N

US

48

D.D.C.

04-CV-02264

Def Dispositive Motion Granted

Y

N

N

N

Y

CI

N

N

US

49

N.D. Ill.

03-CV-06579

.

.

.

.

.

.

.

N

.

50

D.D.C.

04-CV-01864

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

51

E.D.N.Y.

04-CV-00249

Def Dispositive Motion Granted

Y

N

N

N

Y

ER

Y

Y

CB

52

E.D.N.Y.

04-CV-01809

Settlement

Y

N

Y

N

Y

TCD

Y

N

US

53

E.D.N.Y.

04-CV-02743

Settlement

Y

N

Y

N

Y

CI

N

N

US

54

D.D.C.

10-CV-01702

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

55

D.D.C.

07-CV-00984

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

N

Y

FOR

56

D.D.C.

09-CV-02178

Def Dispositive Motion Granted

Y

N

N

N

N

ER

Y

Y

FOR

57

D.D.C.

09-CV-00028

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

58

D.D.C.

06-MC-00145

Def Dispositive Motion Granted

Y

N

N

N

Y

WMD

Y

Y

FOR

59

E.D. Pa.

10-CV-00586

Def Dispositive Motion Granted

Y

N

N

N

Y

TSA

Y

N

US

(Table 2.) Non-Prisoner Appeals Data

No.

Docket

App_Cir

Allege

Locus

Appellant

App_Result

1

12-CV-14766

6th

CWD

US

Plff

Affd

3

10-CV-01492

6th

WCD

CB

Plff

Affd

4

05-CV-03761

7th

TSA

CB

Def

Revd

5

09-CV-09066

2d

CI

US

Plff

Affd

8

04-CV-03950

2d

WCD

US

Plff

Dismiss

10

05-CV-03919

2d

WCD

CB

Def

Dismiss

12

07-CV-13678

6th

CI

US

Def

Revd

14

06-CV-00545

9th

TSA

US

Plff

Partial

18

05-CV-00093

9th

WCD

US

Def

Revd

18

05-CV-00093

9th

WCD

US

Plff

Dismiss

18

05-CV-00093

9th

WCD

US

Def

Dismiss

19

05-CV-01417

4th

ER

FOR

Plff

Affd

20

11-CV-00301

9th

CI

US

Def

Open

20

11-CV-00301

9th

CI

US

Plff

Open

21

12-CV-00823

4th

WMD

FOR

Plff

Affd

22

06-CV-01996

D.C.

WMD

FOR

Plff

Affd

23

08-CV-01677

D.C.

WMD

FOR

Plff

Affd

24

10-CV-00591

9th

WMD

FOR

Cross

DWin

25

08-CV-00035

9th

WMD

US

Def

Revd

26

07-CV-00410

4th

WMD

US

Plff

Affd

28

04-CV-07030

2d

WCD

US

Def

Dismiss

28

04-CV-07030

2d

WCD

US

Plff

Affd

29

04-CV-01427

9th

CI

US

Def

Dismiss

34

05-CV-00024

9th

WCD

CB

Plff

Dismiss

35

02-CV-09802

2d

WCD

US

Plff

Partial

41

06-CV-01720

11th

CI

US

Plff

Affd

42

07-CV-00812

7th

WCD

US

Plff

Open

47

02-CV-01933

5th

WCD

US

Plff

Dismiss

50

04-CV-01864

D.C.

WMD

FOR

Cross

DWin

51

04-CV-00249

2d

ER

CB

Plff

Affd

52

04-CV-01809

2d

TCD

US

Def

Revd

54

10-CV-01702

D.C.

WMD

FOR

Plff

Affd

56

09-CV-02178

D.C.

ER

FOR

Plff

Open

57

09-CV-00028

D.C.

WMD

FOR

Plff

Affd

58

06-MC-00145

D.C.

WMD

FOR

Plff

Affd

59

10-CV-00586

3d

TSA

US

Def

Revd

(80.) Cross-appeals are counted when distinguishing appellants, but excluded regarding all appellants.