concurring in part and dissenting in part:
Because I believe that determining causation as well as damages by inferential statistics instead of individualized proof raises more than “serious questions” of due process, I must dissent from Part IX of the majority opinion. Otherwise, I concur.
Here’s what happened: Hilao’s statistical expert, James Dannemiller, created a computer database of the abuse of each of the 10.059 victims based on what they said in a claim form that assumed the victim’s torture. Although Dannemiller would have said that 384 claims should be examined to achieve generalizability to the larger population of 10.059 victims within 5 percentage points at a 95% confidence level, he decided that only 136 randomly selected claims would be required in light of the “anticipated validity” of the claim forms and testimony at the trial on liability that the number of abuses was about 10,000.
He selected three independent sample sets of 242 (by random selection but eliminating duplicates). Hilao’s counsel then tried to contact and hold hearings or depositions with each of the claimants on the first list, but when attempts to contact a particular claimant proved fruitless, the same number in the next list was used. When the sample results for the first 137 victims proved insufficient to produce the level of sampling precision desired for the project, Hilao’s counsel continued from case 138 to ease 145. Eventually, 124 were completed from list A, 11 from list B, and 2 from list C.
The persons culled through this process went to Manilla to testify at a deposition (which Dannemiller thought was “remarkable”). Dannemiller Narrative Statement, p. 6. He opined that “this random selection method in determining the percentage of valid claims was fair to the Defendant” as “[a] random selection method of a group of 9541 individuals is more accurate than where each individual is contacted.” Id. Further, the statistician observed that “[t]he cost and time required to do 9541 would be overwhelming and not justified when greater precision can and was achieved through sampling.” Id. at 7. Finally, he concluded that “the procedures followed conformed to the standards of inferential statistics and therefore ... the injuries of the 137 claimants examined are representative of the 9541 victims.” Id.
In accordance with the “computer-generated plan developed by James Dannemiller,” the Special Master oversaw the taking of the 137 depositions in the Philippines. In accordance with the district court’s order, the Special Master was to determine “(1) whether the abuse claimed came within one of the definitions, with which the Court charged the jury at the trial held-in Hawaii, of torture, summary execution,- or disappearance; (2) whether the Philippine military or para-military was or were involved in such abuse; and (3) whether the abuse occurred during the period September 1972 through February 1986.” Special Master and Court Appointed Expert’s Recommendations, p. 1. Based on a review of the deposition transcripts of the 137 randomly selected victim claims, and a review of the claims, the Special Master found that 131 were valid within the definitions which the court gave to the jury; the *788Philippine military or para-military were involved in the abuse of the valid claims; and the abuse occurred during the period 1972 through February 1986. As a result, he recommended the amount of compensatory damages to be awarded to the valid 131 claimants, and for the entire class based on the average awards for torture, for summary execution (including lost earnings, which the Special Master determined should be capped at $120,000 per claimant, and which would be determined by the average for the occupation when a witness did not state the amount of income earned), and disappearance (including lost earnings similarly calculated). His report indicates that “for all three categories, moral damages as a proximate result of defendants’ wrongful acts or omissions, Phil. Civ.Code §§ 2216, 2217 were weighed into the compensation.” Id. at 7.
Thus, causation and $766 million compensatory damages for nearly 10,000 claimants rested on the opinion of a statistical expert that the selection method in determining valid claims was fair to the Estate and more accurate than individual testimony; Hilao’s counsel’s contact with the randomly selected victims until they got 137 to be deposed; and the Special Master’s review of transcripts and finding that the selected victims had been tortured, summarily executed or disappeared, that the Philippine military was “involved,” that the abuse occurred during the relevant period, and that moral damages occurred as a proximate result of the Estate’s wrongful acts.
This leaves me “with a profound disquiet,” as Judge Higginbotham put it in In re Fibreboard Corp., 893 F.2d 706, 710 (5th Cir.1990). Although I cannot point to any authority that says so, I cannot believe that a summary review of transcripts of a selected sample of victims who were able to be deposed for the purpose of inferring the type of abuse, by whom it was inflicted, and the amount of damages proximately caused thereby, comports with fundamental notions of due process.
Even in the context of a class action, individual causation and individual damages must still be proved individually. As my colleagues on the Sixth Circuit explained in contrasting generic causation — that the defendant was responsible for a tort which had the capacity to cause the harm alleged — with individual proximate cause and individual damage:
Although such generic causation and individual causation may appear to be inextricably intertwined, the procedural device of the class action permitted the court initially to assess the defendant’s potential liability for its conduct without regard to the individual components of each plaintiff’s injuries. However, from this, point forward, it became the responsibility of each individual plaintiff to show that his or her specific injuries or damages were proximately caused by [the defendant’s conduct]. We cannot emphasize this point strongly enough because generalized proofs will not suffice to prove individual damages. The main problem on review stems from a failure to differentiate between the general and the particular. This is an understandably easy trap to fall into in mass tort litigation. Although many common issues of fact and law will be capable of resolution on a group basis, individual particularized damages still must be proved on an individual basis.
Sterling v. Velsicol Chem. Corp., 855 F.2d 1188, 1200 (6th Cir.1988).
There is little question that Marcos caused tremendous harm to many people, but the question is which people, and how much. That, I think, is a question on which the defendant has a right to due process. If due process in the form of a real prove-up of causation and damages cannot be accomplished because the class is too big or to do so would take too long, then (as the Estate contends) the class is unmanageable and should not have been certified in the first place. As Judge Becker recently wrote for the Third Circuit in declining to certify a 250,000-member class in an asbestos action: “Every decade presents a few great cases that force the judicial system to choose between forging a solution to a major social problem on the one hand, and preserving its institutional values on the other. This is such a case.” Georgine v. Amchem Prod., Inc., 83 F.3d 610, 617 (3d Cir.1996).
*789So is this. I think that due process dictates the choice: a real trial. I therefore dissent.