Vivian Berger Mediator

Home « Mediation, Employment, Civil Rights « Settling Employment Discrimination Lawsuits

Summary Judgment Benchmarks for Settling Employment Discrimination Lawsuits

Hofstra Labor & Employment Law Journal
Fall 2005

Vivian Berger[*], Michael O. Finkelstein[*], and Kenneth Cheung[* ]


The number of employment discrimination lawsuits rose continuously throughout the last three decades of the twentieth century. In the federal courts, such filings grew 2000%, while the docket as a whole increased a mere 125%. [FN1]

In the twelve-month period ending on September 30, 2003, tables put out by the Administrative Office of the U.S. Courts indicate that these actions amounted to slightly more than 8% of all civil cases initiated in that fiscal year (FY).[FN2] Likewise, filings at the EEOC have mounted.[FN3] The two trends are not unrelated; in order to bring an action in court to redress alleged job discrimination, the plaintiff must first exhaust his administrative remedies. Usually, he will do so by bringing his complaint to the EEOC (although in jurisdictions with a similar state or local agency, he may opt to go there instead).

The vast majority of civil lawsuits do not culminate in post-trial judgments; those that are not dismissed before trial very often settle.[FN4] Employment actions are no different. Courts and agencies, concerned about controlling their dockets and using their limited resources efficiently, deem settlement promotion desirable - particularly so in a case category as large as employment discrimination. From a broader vantage point, the public interest in vindicating civil rights laws makes achieving just compromise an aim that transcends individual litigants' goals.

Parties and lawyers entering settlement negotiations must be able to assess the value of the case as accurately as possible. To the extent that the adversaries have congruent views of the worth of the lawsuit, resolution is not only more likely but also more apt to be fair. But in actuality, the parties and the attorneys involved in legal proceedings almost always harbor bias, an understandable tendency to magnify the merits of their own positions.

The plaintiff employee, who has ordinarily resigned or been discharged from his job prior to embarking on litigation, [FN5] is often blinded by serious emotional and financial distress, [FN6] causing an overly optimistic estimate of her chances of prevailing and recovering significant damages. Similarly, defendant employers are often irate at what they regard as the employee's betrayal or extortion, producing the same distorting effect that plaintiffs' strong feelings exert on case evaluation. In a perfectly logical world, the parties' estimates of their probabilities of victory should add up to 100%. In the real world, at the outset of talks, the total may be as great as 150%! Such delusions impede settlement.

Mediation, by contrast, promotes compromise by injecting a neutral third party into a dispute. The growing governmental (including court) usage of alternative dispute resolution (ADR), fueled by legislative mandates or encouragement, has created a large demand for mediation services. [FN7] Many of the cases sent to mediation by judges of the U.S. Courts for the Southern and Eastern Districts of New York (S.D.N.Y. and E.D.N.Y.) are ones alleging job discrimination. [FN8] The EEOC, on its part, also relies heavily on mediation. [FN9]

Unlike courts, administrative agencies, or arbitrators, a mediator does not make factual findings or issue orders; her role is solely to facilitate talks between the parties. Furthermore, mediation participants seeking ways to end controversy are not constrained by governing law - they may agree to anything that is not illegal. As a practical matter, however, the further along a case has proceeded on the litigation track, the more the parties will feel pressured to bargain in the shadow of the law. Otherwise put, in making and weighing demands or offers, they will try to predict what would happen if, instead of settling on specific terms, they went to trial. Also, in "late-stage" employment cases, the employer and the employee (who, as noted above, [FN10] have usually long since parted ways), tend to focus their negotiations mainly or exclusively on money.

In this context, a large part of the mediator's task, while in private caucus with each side, involves "reality testing" the litigants' and lawyers' assumptions and predictions. The mediator assists the parties in estimating the probability of the jury's finding a violation of law and awarding a particular amount of damages. In the job discrimination area, this entails judging how much the employee would likely receive in back pay, front pay, compensation for pain and suffering, and - in rare cases - punitive damages, discounted by the risk of a determination of no liability. In addition, with the mediator's aid, the defendant must calculate the odds that it will have to pay a prevailing plaintiff's attorney's fees, [FN11] as well as take into account the cost of its own counsel and other litigation expenses, which it will incur in any event, win or lose, in mounting its defense. The plaintiff does not have similar concerns. He will not have to pay for opposing counsel and will often not be responsible for his own attorney's fees because plaintiffs' lawyers generally work on a contingency- fee basis. [FN12] This arrangement does, however, give the attorney a stake in the case.

A major part of reality testing in employment actions involves assessing the chance the defendant will succeed in getting summary judgment, in whole or in part, against the plaintiff. In Professor Berger's experience, most employers' counsel say during mediation that they intend to file a "Rule 56" motion [FN13] if the case does not settle; and at least in the Southern and Eastern Districts of New York, a large number of employers do so. [FN14] By contrast, practically no plaintiffs in employment cases make such motions.

The judge's decision to grant or deny summary judgment is a critical juncture - it can make or break a case. Except in the instance of a motion for partial summary judgment, a grant means total defeat for the plaintiff; the case is over unless he brings and wins an appeal, an unlikely prospect. Thus, the mediator has to focus the employee's team on the need to discount anticipated gains by the risk of loss, not just at trial but of trial itself. Even partial summary judgment, if it strikes at the heart of the plaintiff's case, can constitute a major setback for him. For example, if the court dismisses all of the plaintiff's federal claims with prejudice, it is likely to dismiss his supplemental state claims without prejudice. [FN15] In addition, a plaintiff occasionally expresses a wish to go to trial rather than settle even if he understands that his chances of victory are slim, stating that he will be content merely to tell his story to, and be judged by, a jury of his peers. In order to make a rational decision whether to reject offered terms to satisfy this emotional interest, such a litigant plainly requires as accurate an estimate as possible of his odds of talking to a jury at all.

Lastly, even though denial (total or partial) of a motion for summary judgment does not theoretically dispose of the case, as does a grant, practically it may. Defendants often become serious about settling only when confronted by this negative turn of events. Moreover, survival of one or more of the plaintiff's claims greatly raises the settlement value of the case. [FN16] Litigation costs will escalate if the matter is tried. Further, denial of summary judgment forces a defendant to confront squarely the daunting prospect that it may also end up paying substantial damages. Even an employer sanguine about eventual victory knows that juries are unpredictable. In these circumstances, early voluntary resolution may present an attractive prospect to an employer focused on facts rather than emotions. Therefore, a defendant, much like a plaintiff, must be prodded to evaluate its risk of summary judgment carefully.

To be sure, although we have been speaking mainly in terms of mediation, information bearing on the likelihood of summary judgment is also useful to lawyers and clients in standard, unaided negotiations; we hope and expect that our study will assist the employment bar in settlement talks whether or not a neutral party participates. But it does bear emphasis that a key advantage of mediation stems from the presence of an impartial facilitator, who can employ private conferences to explore candidly with each side its vulnerability to a grant or denial of summary judgment (in addition to other litigation perils). Given their bias, both parties tend to view summary judgment through rose-colored glasses until the mediator presses for a more realistic outlook.

Our examination of summary judgment comes fortuitously at a time when scholars, as well as some judges and lawyers, have been drawing attention to the "Vanishing Trial" phenomenon. In addition to articles, symposiums have been devoted to this subject. [FN17] Using data supplied by the Administrative Office, Professor Marc Galanter has shown that in 1962, 11.5% of federal civil cases went to trial; by 2002, the number had dwindled to a meager 1.8%. [FN18]

Interestingly, civil rights cases (a category of which about half fall into the job discrimination area) now comprise more than one-third of civil trials. [FN19] Yet even in this class of cases, the number of trials has declined dramatically: from 19.7% of civil rights dispositions in 1970 to 3.8% in 2000. [FN20]

Where have all the trials gone? The factors generally cited as waxing while trials are waning include settlements, [FN21] ADR [FN22] - and our concern, summary judgment. [FN23] Much of the discussion of summary judgment in this setting consists of bemoaning its major contribution to litigants' (principally, plaintiffs') loss of the constitutionally guaranteed right to a "live" trial, rather than one confined to papers. [FN24] Deploring the loss of trials is basically a response to the "rhetoric of 'explosion' and 'crisis"' fueled by advocates of tort reform. [FN25]

This article, by contrast, takes no position on these issues. Descriptive, not normative, it is designed to aid lawyers and their clients in employment discrimination actions in the district courts of the Second Circuit - with or without a mediator's help - to evaluate the worth of their case with a view toward achieving a workable settlement. We now turn to the study itself.


Our aim was to generate a tool that would help the parties assess the probability of their case's or claim's withstanding a defendant's motion for summary judgment, according to the presence or absence of certain parameters. Based on Professor Berger's experience, those parameters were the following: type of discrimination claim--race, color, national origin, religion, sex, disability, age or retaliation; type of employer, private or public (governmental); and plaintiff's status as represented or pro se. We also created a separate class of "reverse discrimination" complaints, in which we listed claims by males alleging sex discrimination and by Caucasians asserting racial or ethnic bias by members of minorities. However, because color was always linked to a charge of discrimination based on race or national origin, we decided not to code it separately. [FN26]

We made a decision to include only those cases and claims in which the court disposed of a defendant's summary judgment motion. We did so because a favorable decision on such a motion by the court is what attorneys for the defendant commonly predict in mediation, and attorneys for the plaintiff commonly downplay; no one argues about whether the motion will be made. Thus, we did not assemble data to show the rate at which such motions are filed. As an extension of our decision, (i) we counted as a denial any determination denying a summary judgment motion in any respect; and (ii) we treated a partial summary judgment motion that was fully granted as a grant, not a denial. (We have more to say on this coding choice later. [FN27])

The basic statutes protecting employees against discrimination by their employers are: Title VII of the Civil Rights Act of 1964 (Title VII), [FN28] the Age Discrimination in Employment Act of 1967 (ADEA), [FN29] the Americans with Disabilities Act of 1990 (ADA), [FN30] the Equal Pay Act of 1963, [FN31] and the Civil Rights Act of 1866. [FN32] The Rehabilitation Act of 1973 [FN33] provides federal employees with protection equivalent to the ADA. Although a plaintiff will often combine one or more of these claims with other causes of action, [FN34] we confined our study to the abovementioned statutory claims. We did, however, include in our purview causes of action arising under state or city civil rights laws, such as those of New York State, [FN35] New York City, [FN36] and Connecticut, [FN37] which fall under the pendent jurisdiction of the federal courts.

Our sources were two-fold. First, we surveyed all published opinions deciding summary judgment motions by the defendant in employment discrimination lawsuits in the district courts of the Second Circuit for the first three-quarters of calendar year 2001. We relied primarily on Westlaw. [FN38] A less extensive search of cases reported by Lexis yielded virtually no opinions that we had not discovered on Westlaw; hence, we abandoned further research in the Lexis data base. Altogether, we found 191 relevant opinions.

We recognized, of course, that because of the well-known effects of "publication bias" [FN39] this method yields a skewed sample of cases. As Professors Peter Siegelman and John J. Donohue, III have noted, rulings that dispose of a case are more likely to generate opinions and be published than those that do not. [FN40] Yet short of reading hundreds of court files, we could not accurately learn what types of claims plaintiffs had made and courts had ruled on. To avoid the problem of substituting bad data for no data, [FN41] we used the published opinions in ways that we believe were unaffected by this bias. [FN42]

Our second source was the courts' online data base, "Pacer." Among other things, it permits the user to read the docket sheets of all cases in specific categories filed within a certain time frame. Unlike the published opinions, these sheets do not reliably indicate the type of discrimination (race, sex, disability, etc.) alleged. But they do report the parties' represented or pro se status and are indicative of whether a defendant is a public or a private entity. Where we could not determine the latter, we were able to garner this information from the internet.

We searched the sites for the S.D.N.Y. and the E.D.N.Y. under the heading # 442 ("Civil rights: jobs"). For the E.D.N.Y., we used the calendar year 2000 and the first half of 2001, a total of 679 cases. For the S.D.N.Y., a larger data base, we surveyed 610 docket sheets in 2000. [FN43] Looking for cases in which the defendant had made a summary judgment motion that the court had decided, we retrieved 294 out of a sample of 1,289 (679 + 610) filings. Thus, the rate of summary judgment dispositions in cases filed in the two districts for the selected periods comes out to be 22.8%.

When issues arose regarding proper classification of the data, we made a number of categorizing choices. For one thing, the published opinions did not always name the parties' attorneys; in this situation, absent any mention of pro se status, we assumed the plaintiff was represented. In addition, at times, published or Pacer cases revealed that the plaintiff had counsel for only part of the case; we treated the plaintiff's status at the time of the summary judgment motion as salient for coding purposes.

When the only record of a motion's determination was a recommendation by the magistrate judge to whom the matter had been referred, we treated it as final; these recommendations are usually accepted. Sometimes, too, plaintiffs would appeal grants of summary judgment. Such decisions were virtually always affirmed. But where they were not, we naturally treated the appellate decision as dispositive. In a few cases, where defendants made more than a single motion, we recorded the "bottom line": where matters stood after the final determination. In several others, the court denied the motion with leave to re-file later (for instance, where the plaintiff required more discovery). In the absence of relevant subsequent history, we coded the action as a denial.

We focused, moreover, on motions for summary judgment brought by the company "employer," a legal entity. Although the modern federal civil rights laws do not allow for personal liability for individuals who exercise supervisory control over the plaintiff, [FN44] plaintiffs will often name executives or managers as defendants (these are actually "employees"). Represented by the firm's lawyer, they will routinely move for and win dismissal from the case. We did not count this collateral activity as summary judgment grants for defendants; likewise, we ignored motions by non-employers such as unions or employment agencies, although they are also liable for discrimination under Title VII.

Perhaps the most difficult question was how to code the results of motions in cases of more than a single plaintiff; happily, there were very few. We opted for a contextual approach. If the plaintiffs seemed to be virtually identically situated and the court made the same decision as to all, we counted the motion only once. If, by contrast, the plaintiffs had fewer commonalities - in outcome or in circumstances - we recorded data for as many motions as there were plaintiffs. In one such instance, arising on Pacer, we had to consult the court file to get the pertinent details. [FN45]

A couple of remaining comments concern solely the Pacer cases. We stress that we recorded data for summary judgment dispositions, not for all motions made. Thus, if a motion was pending when the case settled, we ignored it. Sometimes, however, though not issuing a formal ruling, the judge behaved in a way we treated as a denial, albeit implicit. For example, in several instances, the case went to trial without the court's having acted on the motion. [FN46] In one instance, the court rejected the defendant's express application to have the motion heard before trial. Such actions or inaction, charitably viewed, might reflect the judge's belief that uncertainty, coupled with the looming specter of trial, would promote settlement, thereby conserving both the parties' and court's resources.


Out of our Pacer filing sample there were 294 cases in which a summary judgment motion had been made and disposed of by the court. Of those 294 cases there were 107 with motions that were denied, for a rate of 36.4%. Because of publication bias one would expect that in the universe of cases represented by our Pacer sample, the denial rate (full or partial denial of the motion) would be higher than in the published cases. [FN47] Our data do reflect that bias, as appears in Table 1 below.


Table 1
Pacer vs. published cases: summary judgment denial rates for represented and pro se cases combined, private and public employers combined, and all districts combined.*
Sample # denied/ # total % denied
Pacer 107/294 36.4%
Published 59/191 30.9%
*In this and successive tables the districts for Pacer are the Southern and Eastern Districts of New York and for published cases are the Southern, Eastern, Northern, and Western Districts of New York, and the Districts of Connecticut and Vermont.


The bias effect is relatively small - 5.5 percentage points - but part of the effect is concealed by the larger number and proportion of pro se plaintiffs in the Pacer cases. [FN48] One would expect to see relatively fewer pro se matters generating published opinions: because of unrepresented litigants' disadvantage in articulating legal arguments, their cases are unlikely to produce opinions of interest to anyone except the parties. Thus, when one compares only represented cases, the difference becomes larger - 11.3 percentage points - as shown in Table 2 below. The table shows that summary judgment motions are denied in almost half (46.4%) of represented cases.

Table 2
Represented Pacer and published cases: summary judgment denial rates for private and public employer cases combined; all districts combined.
# denied/ # total
% denied
* A 95% confidence interval for this 11.3 percentage point difference is 0.5 percentage points to 22.2 percentage points.

The pro se plaintiffs in both Pacer and published cases have summary judgment denial rates much lower than those of represented plaintiffs. See Table 3 below. This is hardly surprising. Judges often purport to read pro se papers with a thumb on the scale for the plaintiff, [FN49] and the Rules of the U.S. District Courts for the Southern and Eastern Districts of New York require a represented movant for summary judgment to serve on a pro se opponent a "Notice to Pro Se Litigants Opposing Summary Judgment." However, such "band aid" devices cannot realistically level the playing field for the unrepresented party. It is likely, moreover, that some, perhaps most of the pro se plaintiffs tried to hire counsel but were turned down because the lawyer believed they had poor chances of victory. Hence, pro se cases start out weak and become weaker on account of the lack of professional advocacy. Because of the distorting effect of pro se status, the rest of our analysis, except for Table 4, is confined to represented cases.


Table 3
Pro se Pacer and published cases: summary judgment denial rates; private and public employer cases combined; all districts combined.
# denied/ # total
% denied
* A 95% confidence interval for this 11.3 percentage point difference is 0.5 percentage points to 22.2 percentage points.


The foregoing tables have combined results for the Eastern and Southern Districts and for public (governmental) and private employers. In represented cases, the Eastern District (46/98 = 46.9%) has virtually the same denial rate as the Southern District (45/98 = 46.4%), and so we decided to amalgamate them, arriving thus at the 91/196 = 46.4% figure in Table 2. Public employers have higher motion-denial rates than private employers, as shown in Table 4 below, and the difference is of marginal statistical significance. It is tempting to speculate about the cause for this discrepancy. A colleague of ours thought that lawyers in the private sector might furnish, on the whole, more skilled and zealous representation. This possibility cannot be excluded. But in Professor Berger's experience, at least attorneys for the New York City Law Department as well as Assistant U.S. Attorneys provide service (in mediation, at any rate) that is generally on a par with their private counterparts'. In sum, we are not confident that we know of any reason for the higher denial rate in cases against public employers; thus, despite the finding of marginal significance, the results may well be due to chance. [FN50] Accordingly, except for Table 4, we have combined both types of employers.


Table 4
Pacer Cases: summary judgment denial rates by representation and type of employer; Southern and Eastern districts combined.
Type of Employer
Represented Cases
Pro Se Cases
# denied/
# total
% denied # denied/
# total
% denied
Private 47/117 40.2%* 8/55 14.5%
Public 44/79 55.7%* 8/43 18.6%
* The P-value for this 15.5 percentage point difference is 0.041.

We wanted to examine the consistency of our data with those presented in other studies. Because most other studies focused on the "Vanishing Trial" phenomenon, they considered trial rates or total pre-trial terminations averaged across all district courts, or groups of districts not confined to a single circuit. In addition, they usually combined broader collections of types of cases and made no adjustment for pro se status. One study using Administrative Office data is of particular interest because it is relatively recent and focuses on employment discrimination cases. [FN51] Professors Kevin M. Clermont and Stewart J. Schwab report that in FY 2000 the rate of "nontrial adjudications" in employment discrimination cases for all district courts combined was 20%. [FN52] The fact that this is higher than ours (14.5% [FN53]) may be due in part to (i) their categorization of "nontrial adjudications" as encompassing all types of motions (for example, motions to dismiss as well as for summary judgment), whereas we consider only summary judgment motions and, perhaps more significantly, (ii) their inclusion of all districts, whereas we include only the S.D.N.Y. and E.D.N.Y. Notably, the former is reputed to be one of three districts with the most restrictive approach to granting summary judgment. [FN54]

Claim Survival by Type of Claim

Are some types of claims more likely to survive summary judgment motions than others? To address this question, we examined summary judgment denial rates in the published cases by claim type. (Recall that the Pacer cases do not reliably indicate the kind of discrimination involved. [FN55]) The rates of claim survival are all remarkably uniform - between 24.3% and 31.0% - with the notable exception of sex claims, which have a survival rate of 41.7% (see Table 5, second column). The survival rates, adjusted for publication bias, are shown in the last column; the method of adjustment is explained below. [FN56]

Table 5
Represented published cases: claim survival rates by type of claim; public and private employers combined; all districts combined.
# Survived /
# Total
% Survival
Adjusted % Survival
* National origin (5/17) + religion (2/7) + reverse discrimination (2/5).


The higher survival rate for sex claims (although not statistically significant because of small sample size) appears to be due to the inclusion of sexual harassment claims. We define these as accusations of unwelcome sexual conduct, not solely charges that male co-workers or supervisors created a generally hostile environment for a female plaintiff (as, for example, by making demeaning remarks about women or depriving them of opportunities given to men). For ease of description, we refer to the latter as "gender," as opposed to "sexual harassment," claims. [FN57] Of the total of 48 represented sex cases in our sample, 27 involved solely gender discrimination; for these the survival rate was 9/27 = 33.3%. By contrast, for the 21 sexual harassment cases, the survival rate was 11/21 = 52%.

We have some thoughts about why complainants alleging sexual harassment do better than others in resisting summary judgment. For one thing, such cases frequently depend heavily on credibility assessments. When the purported misconduct took place behind closed doors, the historical facts raise classic "he said, she said" problems, not susceptible to resolution on papers alone. In addition, even in the absence of material disputes over primary facts, a finding of sexual harassment depends on inferences drawn about the alleged harasser's behavior and its effect upon the plaintiff. Specifically, the focus is on whether a reasonable person would have found the environment hostile and whether the victim, in fact, perceived it that way. [FN58]

The latter issue plainly raises subtle questions about the plaintiff's subjective feelings, which are ill-suited to determination on summary judgment. Beyond that, however, as District Judge Jack B. Weinstein suggested in Gallagher v. Delaney, [FN59] the harassment inquiry's objective part also does not lend itself to pretrial judicial resolution. Judge Weinstein wrote:

A federal judge is not in the best position to define the current sexual tenor of American cultures in their many manifestations. . . .

. . .Today, while gender relations in the workplace are rapidly evolving and views of what is appropriate behavior are diverse and shifting, a jury made up of a cross-section of our heterogenous [sic] communities provides the appropriate institution for deciding whether borderline situations should be characterized as sexual harassment . . . . [FN60]

One might ask whether Gallagher has expressly motivated district courts to deny summary judgment motions in cases of alleged sexual harassment. Research reveals that 15 of our published opinions cited Gallagher. But only three opinions used it in the context of the need for jury determination of pertinent issues (rather than, say, as support for a black-letter rule of law), [FN61] and one of these granted the motion. [FN62] This is not to say, however, that the same reasons that impelled the Gallagher court to view sexual harassment claims as poor candidates for summary judgment have not influenced district judges to come to the same conclusion themselves.

Moreover, allegations of sexual misconduct, unlike charges of wrongful dismissal, failure to promote, or other forms of discrimination based on gender, cannot be countered with the typical defense that the plaintiff's performance was unsatisfactory. Nor, to be sure, can allegations of hostile environment founded on race or national origin, for example. Yet these typically do not raise "borderline" issues of characterization, credibility contests, or complex questions of victim perception to the same degree as cases involving sexual harassment. [FN63]

A second point of particular interest is that retaliation - the largest category - does not have a higher survival rate and, in fact, has a slightly below-average rate. Received wisdom is that retaliation claims have higher survival rates. This makes sense. A plaintiff can prevail on retaliation without proving the validity of the underlying claim of discrimination: he need only show that he engaged in protected activity known to the employer, that he suffered an adverse employment action, and that the former caused the latter. [FN64] Many cases present the pattern of a bias complaint by the employee (which is protected), followed by his termination. If the time interval between the two is relatively brief, a jury will ordinarily be permitted to infer causation. [FN65] Hence, the relatively lower rate we found for retaliation claims may be due to sampling error. [FN66]

Finally, it bears brief mention that disability claims came in lowest in terms of survival (though, again, not by much). ADA claims in general do poorly in the courts [FN67] because of restrictive interpretations given the statute by the U.S. Supreme Court. [FN68] We would expect that the rate for this category would have been even lower if not for the fact that New York State and New York City have laws defining "disability" more broadly than the ADA; [FN69] knowledgeable plaintiffs routinely include such pendent claims in their complaints. [FN70]

Adjusting Claim Survival Rates for Publication Bias

The figures in the second column of Table 5 apply only to claims in published cases, leaving us with publication bias. To adjust for this effect, imagine there is only a single claim made in each case. In that event, the adjustment for publication bias for claims would be the same as the adjustment for cases. This would be 46.4% / 35.1% = 1.32 (see Table 2). Thus, the overall claim survival rate would rise by a factor of 1.32, from 29.5% to 38.9% (see Table 5). If publication bias affects different kinds of claims equally, the survival percentages for each type of claim would increase by 1.32, with the results shown in the last column of Table 5. Making this adjustment, the survival rate for sex discrimination claims rises to 55%, with the rest of the claim types ranging between 32.1% and 40.9%.

We justify applying this average adjustment rate to different types of claims because the impetus for opinion writing and publication would appear to be largely unrelated to the type of claim. [FN71] That seems a reasonable assumption, at least when survival rates for different kinds of claims are fairly close, as they are here. The one exception is sex discrimination: since summary judgment is denied more often in these cases than in other types, and since denials of summary judgment are more likely to be unpublished than grants, sex discrimination cases may be over-represented in unpublished cases. If this is so, the adjustment factor would be smaller than the 1.32 figure used in Table 5. We note that even if a systemic difference existed, the effect would be dampened by the fact that a substantial portion of the Pacer cases had published opinions. [FN72]

Adjustment for Multiple Claim Cases

The adjustment for publication bias was based on case-survival figures, and was equal to the adjustment for claim survival on the assumption of one claim per case. But since there are many cases with multiple claims, we must now consider whether a further adjustment for that factor is warranted. [FN73] We refer to the factor by which case survival rates must be multiplied to arrive at claim survival rates as the "claim-adjustment" factor. [FN74]

When there are multiple-claim cases, the rate of case survival will be higher than the rate of claim survival because the motion fails if any challenged claim survives. Thus, the claim-adjustment factor should be less than 1. Our published case data do demonstrate this result: the claim survival rate is 84% of the case survival rate. [FN75] The size of the claim-adjustment factor is a function of the way claims are distributed among cases. This is not known for the unpublished cases, but we think it reasonable to assume that the distribution of claims across cases is the same for unpublished as for published cases. In that event, the claim-adjustment factor would be the same for both groups of cases and would cancel out in the adjustment ratio. [FN76] Hence, we made no change in the ratio.

Other Sources of Bias

In addition to publication bias and case effect there are other possible sources of bias. First, the Pacer cases are only from the S.D.N.Y. and the E.D.N.Y., while the published cases come from all the district courts in the Second Circuit. However, when published cases are confined to the Southern and Eastern Districts the results are virtually unchanged: the survival rate for represented cases goes from 35.1% to 34.6% (37/107). The survival rates for claims are also not greatly changed. For example, in the case of sex discrimination, the rates of survival are almost the same for the two groups of districts separately, so their combined figure fairly represents each. [FN77] Second, the Pacer data are cases filed in 2000 and 2001 with summary judgment decision dates ending in June 2005; the published cases have decision dates in 2000-2001. The time frames are, therefore, slightly different. However, they substantially overlap, and we know of no intervening developments in the law that would have affected survival rates. Third, not all the Pacer cases were resolved in our study window, creating the possibility of bias by excluding pending cases. This "tail" problem is endemic in legal studies, and Mr. Finkelstein and his colleagues have dealt with it elsewhere. [FN78] But the problem does not appear to be serious here: after a follow-up there were only 14 pending Pacer cases compared with 294 completed cases. Accordingly, the excluded cases were less than 5% of the total.

Fourth, one may question our decision to treat partial summary judgment motions as granted (as long as they are fully granted) even when the case survives for the claims not challenged. We recognize the problem, but treating partial summary judgment motions as denials with regard to the unchallenged claims would be somewhat inconsistent with our decision to look at survival rates only for cases in which a motion was made. Either treatment has its drawbacks. Fortunately, there are only two published cases in our data in which partial summary judgment motions were fully granted, so the issue is not material.

The Appraisal of Cases

Assuming that our estimates of average survival probabilities for the various types of claims are good starting points for evaluating the claims, how should a mediator or negotiator assess the probability of some claim surviving? When there is a single claim, the adjusted percentages in the last column of Table 5 provide a fair starting point for estimating survival probabilities. When there are multiple claims, the problem is more difficult. Because the fates of claims are highly intertwined, in most cases it would be inappropriate to compute the probability of a claim surviving as if claim survivals were independent events. The alternative is to estimate survival rates for different groupings of claims. Unfortunately, there are many possible groupings, and we do not have the large database that would be required for that purpose. [FN79]

In our experience most cases have a claim that is the gravamen of the action. As a result, a mediator or negotiator should probably be guided by the probability of survival for that type of claim without adding much or anything for other claims' survival probabilities. The one exception might be where the second claim is for retaliation, and in such cases some addition may be appropriate. [FN80]


Our Pacer data provide an estimate that, in represented cases, almost half of summary judgment motions are at least partly denied. Based on our experience, this is probably closer to the plaintiff's estimate of survival chances than most defendants'. By contrast, where the plaintiff is pro se, denials occur in only about 15% of the cases. The difference presumably arises both from the fact that the case was weak (no lawyer, if any was asked, would take it) and the plaintiff's disadvantage of lacking a lawyer to fight the motion. [FN81] Nothing else is as predictive of the result on summary judgment as the status, represented or not, of the plaintiff. We also found a somewhat higher rate of summary judgment denials when the defendant is a public, as opposed to a private, employer. The reason for this difference remains unclear.

When we compare our Pacer cases with our published cases, there is evidence of publication bias: the former have a higher survival rate. This is an expected result because a judge granting summary judgment will almost surely write an opinion, but will not necessarily do so when denying the motion.