A group of defendants, currently engaged in a highly complex mass tort litigation, have filed a mandamus petition asking us to disqualify the judge and to declare a mistrial because two of the judge’s law clerks have brothers who represent plaintiffs. After reviewing the special circumstances of this litigation, we conclude that it would not be proper for us to provide the relief these defendants seek. We therefore deny the defendants’ petition for mandamus.
I.
Background
On New Year’s Eve 1986 a fire in the San Juan Dupont Plaza Hotel in San Juan, Puerto Rico killed 97 people and injured several hundred others. Within a few months over 2,300 plaintiffs filed numerous actions seeking a total of about $1.8 billion from more than 200 defendants. The Judicial Panel for Multidistrict Litigation transferred these related cases from all over the United States to San Juan, Puerto Rico, where Judge Raymond Acosta has since managed, and brought to trial, what the trade press has described as “the nation’s largest mass disaster litigation.”
Judge Acosta has played an active role in trying to obtain a resolution of the litigation. He developed and promulgated a 252-page case management order. He imposed and enforced strict discovery deadlines. He appointed a Plaintiffs’ Steering Committee. At his request the Chief Justice of the United States appointed U.S. District Court Judge Louis C. Bechtle of Philadelphia a “settlement judge,” permitting the parties to try to settle their claims before that judge while they simultaneously prepared for trial in Judge Acosta’s court.
Judge Acosta’s management orders provided for several separate trials, each dealing with different issues common to many of the individual cases. A “Phase One” trial permitted the plaintiffs to litigate against the hotel’s many corporate owners (and related insurers) to determine whether the plaintiffs could “pierce the corporate veil” of the corporation that directly owned the hotel (and had little insurance) in order *969 to make more distant (and ultimate) owners liable. In the “Phase Two” trial now before us, the plaintiffs are litigating against manufacturers of products used in the hotel and the providers of services to the hotel, claiming that defects in those products and services led to unnecessary death and injury. Trials of later phases will involve apportionment of negligence, various cross-claims and third party complaints, and ultimately the hotel’s racketeering claim against the Teamsters Union, whose members, the hotel says, set the fire.
After about two years of pretrial proceedings that involved the collection and storage of more than 3,000,000 documents, the taking of 2,200 depositions, 200 pretrial orders, 500 “margin orders,” 12,000 docket entries, and the construction of a specially designed courtroom, the consolidated cases were ready for the Phase One “corporate veil” trial. It began on March 12, 1989. Within two months the hotel defendants and the plaintiffs, negotiating in Philadelphia, reached a settlement, calling for the hotel defendants to pay about $85 million, approximately 35% of the $248 million “global value” that Judge Bechtle said he would place upon the claims against them. On May 12, 1989, Judge Acosta stayed the Phase One trial in light of the settlement.
Judge Acosta then proceeded to the Phase Two “suppliers” trial (against 90 of the hotel’s product and service suppliers); on June 27,1989, jury selection in that trial began. Six weeks later, after opening arguments and the beginning of the trial, 40 of 78 remaining Phase Two defendants asked Judge Acosta to recuse himself and to declare a mistrial. They pointed out that one of the judge’s law clerks who was working on the case, Richard Graffam, has a brother William who is a named partner in a San Juan law firm that represents 58 of the 2,300 plaintiffs; that one of the members of that firm (David Indiano) is a member of the Plaintiffs’ Steering Committee; and that William himself had appeared at the counsel table twice during the Phase Two trial. They also pointed out that another clerk who was working on the case, Vilma Vila, has a brother Raul who is a member of a law firm that represents the hotel corporation, a corporation which, like the plaintiffs, has an interest in finding the suppliers liable for some of the damages. (The Phase One settlement agreement explicitly provides that the $85 million payable by the hotel defendants will be reduced by 50% of the amounts plaintiffs receive from other defendants, up to a total reduction of $7 million.) They argued that, in light of these relationships, Judge Acosta’s “impartiality might reasonably be questioned,” 28 U.S.C. § 455(a); hence, the law required him to “disqualify himself.” Id.
Judge Acosta denied the defendants’ motions. The forty defendants (joined by several others) now ask us to issue a writ of mandamus requiring Judge Acosta to declare a mistrial in Phase Two and to disqualify himself from further participation.
See In re United States,
II.
The Relevant Law
The relevant statute, 28 U.S.C. § 455(a), says:
Any justice, judge, or magistrate of the United States shall disqualify himself in any proceeding in which his impartiality might reasonably be questioned.
*970
28 U.S.C. § 455(a). We draw our legal standards for review of a district judge’s decision not to disqualify himself from an analogous case,
In re United States,
III.
The Law Applied
Given these standards, the district court’s determination, and the record before us, we cannot find the district court’s decision unlawful. Were this an ordinary ease, involving a law clerk whose brother (or whose brother’s firm) represented a party, the legal question of whether a judge must disqualify himself would at least be a close one. On the one hand, the relationship at issue, “brother of a clerk,” is a close relationship; that relationship itself creates a fairly close relation between clerk and party, at least where the brother himself (and not simply someone else in a large firm) represents a party in court; and the relationship of clerk to judge itself is close enough that one might find an appearance of undue influence.
See, e.g., Parker v. Connors Steel Co.,
On the other hand, the conflict is relatively weak and remote. Assuming the family relationship raises a slight cloud, few knowledgeable people would expect that it would ordinarily cause most law clerks to actually commit the serious ethical breach of seeking to influence a judge improperly. Both bench and bar recognize, moreover, that judges, not law clerks, make the decisions. The statute itself speaks of “justice[s], judge[s], or magistrate^],” not clerks. See 28 U.S.C. § 455(a). And judges are fully capable (and believed by reasonable members of the public to be fully capable) of taking account of whatever “bias” having a brother in a plaintiff’s law firm might bring to a clerk. Furthermore, individual judges must enjoy considerable independence in deciding precisely how to use their law clerks, for “bureaucratizing” the law clerk system with too many rules and regulations or too much appellate court supervision would threaten the flexibility that permits that system to speed the resolution of cases. At the same time, it is not necessarily easy, from an administrative perspective, for a judge to find and to deal in advance with all “suspect” relationships of staff and parties, particularly in a legal world where “the country’s 25 biggest law firms together now employ more than 10,-000 lawyers.” New York Times, November 22, 1989, at A22, col. 4.
The ease before us, moreover, is not an ordinary case. It is large, complex, and time consuming; it involves hundreds of lawyers, thousands of parties, and hundreds of millions of dollars in potential damages. It is a case where the need for judicial management by the district court counsels caution prior to intervention by an appellate court. And it is a ease of a sort that the Supreme Court has singled out for somewhat special treatment.
See Liljeberg v. Health Services Acquisition Corp.,
1. Section 455’s Appearance of Impartiality. In addition to the general factors already mentioned, three special features of this case militate against our finding an appearance of partiality that requires- disqualification. First, other things being equal, the more common a potentially biasing circumstance and the less easily avoidable it seems, the less that circumstance will appear to a knowledgeable observer as a sign of partiality. If only a handful of judges paid federal income tax, for example, one might wonder about the appearance of one such judge deciding a tax case that could produce lower rates; when nearly all judges pay taxes, the appearance does not seem significant. Similarly, in this case, the federal bar in Puerto Rico is small; fewer than 500 lawyers practice in the federal district court. The number of Puerto Rican lawyers involved in this case is large; 51 Puerto Rican law firms (including 12 of the largest 20) represent defendants; at least 48 Puerto Rican lawyers represent plaintiffs. The risk that a law clerk, or some other staff member, will have a brother or sister or some other family member involved in this case is a likely concomitant of trying such a large case in such a small district. We do not say that the relationship is unavoidable; we simply say that it arises out of a common circumstance. And, a knowledgeable objective observer is therefore more likely to see the relation as implicit in the special circumstances rather than as an odd coin *972 cidence the failure to avoid which might suggest bias.
Second, other things being equal, the greater the extent to which the potentially-disqualifying circumstance facilitates the just and efficient resolution of a case, the less likely a knowledgeable observer will consider it a sign of judicial partiality. In this case, the effective management of this highly complex litigation benefited from, if it did not absolutely require, the services of the particular law clerks in question. Graf-fam and Vila are not law clerks hired fresh from law school. Rather, they are career clerks who were already working for Judge Acosta when the hotel litigation began in January of 1987. The record suggests that their experience and participation were highly useful in bringing Phase One to trial, smoothly, and with unusual speed. In other words, the court’s need for the clerks was somewhat special and likely to have been seen as such.
Third, the parties’ own words and deeds may help determine the extent to which a knowledgeable observer would see, in a particular circumstance, a sign of partiality. Here, the record suggests that the district court reasonably believed that the parties, certainly by the time of the Phase Two trial, were aware, or should have been aware, of the relationships involved. The law clerks were present at nearly all discovery conferences and every day at trial; the press identified them by name, and carried a picture of Richard Graffam together with Judge Acosta; they participated actively in many of the proceedings; during the Phase One jury selection, Richard Graffam’s name and Vilma Vila’s name were mentioned to the jury several times each day. The Puerto Rican legal community, particularly the federal bar, is small enough for active members to know each other, at least by reputation; there is apparently only one Graffam family in Puerto Rico; William Graffam is a prominent lawyer whose name has appeared often on documents associated with this case; indeed, one lawyer for defendants says in a recusal affidavit that he knew that Richard and William Graffam were related before the beginning of the hotel litigation. Yet, for more than two years during which Richard Graffam and Vilma Vila participated in the proceedings, no one objected. This fact suggests that, until recently, those associated with the case did not see the “sibling relationship” as carrying with it a threat to the impartiality of the proceedings; and that fact helps to support the district court’s conclusion that a “knowledgeable” objective observer would not have seen in the relationship a significant indication of partiality.
These special features of this case make it readily distinguishable from
Parker v. Connors Steel Co.,
2.
The Proper Remedy.
Even if we reached the opposite conclusion about the appearance of partiality, we should still deny defendants’ mandamus petition for reasons related to remedy. The petition asks for certain relief that is related to the past, namely the declaration of a mistrial; it also asks for certain relief that is related to the future, namely an order requiring disqualification. As to the future, the appropriate order from this court would require Judge Acosta to proceed without the clerks, not to disqualify himself. “If a clerk has a possible conflict of interest, it is the clerk, not the judge, who must be dis
*973
qualified.”
Hunt v. American Bank & Trust Co.,
We shall not now insist, however, that Judge Acosta disqualify his clerks for the future, for proper management of this enormously complex case is of great importance, we are uncertain about the extent to which these particular clerks are needed at this stage of the proceedings, and we would not risk disruption without such knowledge. We note, however, that one of the circumstances we relied upon in our discussion of impartiality — the absence of objection by the parties — can no longer be said to characterize the situation. We therefore point out to Judge Acosta that he may wish to reconsider the continued use of clerks with a “sibling relationship.”
As to relief for the past, the Supreme Court has said, in respect to judicial actions already taken, that the disqualification statute, 28 U.S.C. § 455, “neither prescribes nor prohibits any particular remedy for a violation of” the duty that the statute imposes.
Liljeberg,
Were our conclusion about the appearance of partiality different, the factors that the Supreme Court mentions, as applied to the case before us, would nonetheless prevent us from setting aside any previous judicial action. Retrying Phase Two would create a significant risk of injustice to the parties, in that doing so would mean considerable repetition. By September 14, 1989, second phase preparation and trial had involved the entry of 2,210 items on the docket, the admission of 699 exhibits, the issuance of 143 orders, and the examination of 57 witnesses. Moreover, interrelationships between the different phases of the case and between settlement procedures and trial, the care with which the pretrial orders, the discovery, and the trial preparations have been crafted and organized, and the way in which those orders have been meticulously enforced, lead us to fear that retrial could mean, not only the unfairness inherent in the loss of time and money, but also the greater injustice inherent in putting the settlement agreements at risk.
Nor do we believe that denial of the particular retrospective relief requested would produce injustice in other cases. The petitioners have not alleged that actual bias has infected any findings or rulings. They have not pointed to any specific findings or rulings already made in the Phase Two litigation that are incurable or could have preclusive effect in some other action.
Further, for reasons set forth earlier, see pp. 970-972, supra, we would not believe (even were the “appearance of partiality” line drawn differently) that the relevant public’s confidence in the judiciary would be seriously undermined were no mistrial declared.
Finally, few equitable considerations favor the parties who have brought this petition. All but one of the 49 petitioners are represented by local counsel, yet only 6 local counsel, representing 24 of the petitioning defendants, have produced sworn statements that they did not know of the “sibling relationships” until after the Phase Two trial began. (Of course, some “stateside” counsel have produced such statements, which is pretty much beside the point.) Given the opportunities for the parties now before us to learn of this “relationship,” and to ask Judge Acosta to find new clerks well before the present trial, we cannot find many considerations of justice that argue for declaration of a mistrial.
In sum, the remedial considerations mentioned by the Supreme Court in Liljeberg weigh heavily against granting the retrospective relief requested.
*974 For both sets of reasons — those related to the merits of the “impartiality” question and those related to remedy — the petition for mandamus is
Denied.