The opinion of the court was delivered by

KING, P.J.A.D.

Plaintiffs, electrical utility customers who experienced power outages on the 4th of July 1999 weekend brought this class action for damages for failure to provide service. They also asserted claims for consumer fraud, negligence and breach of contract. This is an interlocutory appeal on leave granted from the order of decertification of plaintiffs' class and the denial of their motion to admit expert testimony on class-wide damages. R. 2:2-3(b).

We affirm the decertification of the class and the rejection of the proffered proof of class-wide damages. We remand for certification of a more limited class of customers whose outages directly resulted from alleged negligence in delaying the replacement of transformers at the Red Bank substation.

I

Plaintiffs are residential and business customers of defendants GPU, Inc., and related affiliates (GPU or the utility), providers of electrical service. On July 20, 1999 the Tzannetakis plaintiffs filed a class action complaint in the Superior Court, Law Division, Monmouth County, alleging damages resulting from power failures in July 1999. On July 22, 1999 Muise filed a class action complaint in the same court against the same defendants, with similar allegations.

On August 4, 1999 Muise filed a motion for class certification. In September 1999 the Tzannetakis plaintiffs also moved for class certification. On October 8 consolidation was ordered. In an October 12, 1999 hearing Judge Chaiet granted plaintiffs' motions for class certification and denied defendants' motion to dismiss in favor of deferring to the primary jurisdiction of the Board of Public Utilities (the Board).

On January 11, 2000 we granted defendants' motion for leave to appeal from the denial of their motion to dismiss, "in order to determine the proper forum for [plaintiffs'] claims." Muise v. GPU, Inc., 332 N.J. Super. 140, 148 (App.Div. 2000). Defendants did not seek leave to appeal from the class certification because they wanted to develop the factual record. We concluded that the Law Division judge properly retained jurisdiction, rather than deferring to the Board. Id. at 146. We stated, however: "Future developments may require reference of certain issues to the Board, but not at present." Ibid.

At a hearing on May 24, 2001 Judge Uhrmacher denied defendants' motion to decertify the class. On January 28, 2002 she heard arguments on defendants' motion for partial summary judgment. In an opinion issued on August 14, 2002 Judge Uhrmacher granted defendants' motion for partial summary judgment, dismissing with prejudice plaintiffs' claims for fraud, negligent misrepresentation, violation of the Consumer Fraud Act (CFA), N.J.S.A. 56:8-1 to -20, and strict product liability.

On November 8, 2002 defendants filed a renewed motion to decertify the class. On December 17, 2002 plaintiffs filed a notice of motion for a declaration of admissibility of their experts' class-wide damage model. On November 10, 2003 Judge Perri issued her opinion denying plaintiffs' motion to permit expert testimony on their proposed survey model for proof of class-wide damages, and decertifying the class. On January 5, 2004 we granted plaintiffs' motion for leave to appeal this ruling.

II

Plaintiffs alleged damages resulting from electrical power interruptions, deficiencies or failure, and blackouts without notice, during a heat wave from July 4-10, 1999. The Tzannetakis plaintiffs specified damages including spoilage of food; damage to computers, appliances and other electronic equipment; inconvenience, emotional and physical pain and suffering and loss of use of electrically-powered equipment; loss of business income; and other consequential and incidental damages including the cost of alternative sources of power.

Planning engineer Andrew Zulkosky for GPU said it provides electrical service to about one million customers in two noncontiguous regions of New Jersey, separated by a territory served by Public Service Electric Gas Company (PSEG). The North Region includes all or part of Sussex, Morris, Passaic, Essex, Warren, Hunterdon, Union, Somerset and Mercer Counties, and the Central Region includes all or part of Monmouth, Middlesex, Ocean, Mercer and Burlington Counties.

The electrical infrastructure has three main levels: the interstate transmission system (originating at the power plants), the subtransmission system (here, two regions, functioning to reduce voltage and circulate power through a series of looped substations), and local distribution systems (here, more than 1000 circuits, functioning to further reduce voltage and distribute power radially to a particular geographic area). Voltage is again reduced at over 160,000 pole-top or pad-mounted transformers in the two energy regions before power is fed from each to a small number of customers (usually three to ten).

All levels of the electrical infrastructure are equipped with devices to protect individuals from harm and to protect equipment from damage; examples are circuit breakers, protective relays and fuses. Some of the devices operate automatically and some are used by system operators to avoid overloads when necessary to protect the system. Planning engineers attempt to forecast system conditions several years ahead to assure sufficient capacity. In a typical instance of widespread service interruptions, usually caused by severe weather, there are numerous "trouble locations" where those interruptions occur.

According to the Board, during the week of July 4, 1999 hot weather combined with the holiday created unprecedented demand for power. Defendants concede that during the relevant period about 250,000 of its customers lost power for periods ranging from a few minutes to more than fifteen hours. Defendants attribute these interruptions to at least forty-four distinct sources, including equipment failures, deliberate circuit shutoffs to permit access for maintenance or fire fighting, the triggering of protective devices, lightning and automobile accidents.

The parties do not seem to dispute the Law Division judge's synopsis of the power outages:

The evidence before the court, which is voluminous and will only be set forth here as a sampling, indicates that the first significant power outage occurred on July 4, 1999, with a "phase" or burn-out of a line in the Mantoloking substation which interrupted power to the customers served by that substation. Within the next 12-hour period, line repairs were required in Manasquan, which required the system operators to de-energize the system while the work was done. At the Cedar Bridge substation, power had to be interrupted for work to be performed on a messenger wire. On the evening of July 6, 1999, the operation of protective devices at the Manasquan substation affected 1,500 customers.

In the late evening hours of July 5, 1999, a bushing on the Bank 2 transformer at the Red Bank substation failed. The loss of this transformer, coupled with the unusually high demand occasioned by the heat, placed additional stress on the other transformers at the substation. On July 6, 1999, the Bank 1 transformer began overheating and service was manually interrupted to several distribution circuits in order to lessen the load on the transformers. These manual interruptions affected approximately 10,000 customers for periods ranging from 2.5 hours to 8 hours. On July 6, 1999, the Bank 8 transformer also overheated, and the Bank 7 transformer was automatically taken out of service by protective devices. These events affected approximately 70,000 customers.

During this time, GPU also implemented rolling blackouts in order to limit the load on Banks 7 and 8. The rolling blackouts continued through July 7, 1999 and affected approximately 46,000 customers. Typical service interruptions lasted for approximately four hours. On July 8, 1999, rolling blackouts continued but only lasted for 2 hours. At approximately 1:30 p.m. on July 8, 1999, the rolling blackouts ceased when there was a break in the heat and the load on the transformers lessened, allowing the two transformers to handle the load.

Other service interruptions occurred throughout the North and Central regions during this time. GPU has identified no less than 38 instances of entire circuit interruptions, the reasons for which included regulator failure, circuit breaker failure, underground feeder cable failure, line burn outs, operation of over-current relay protectors, circuit de-energization, tripping of transformer breakers and carpole accidents. Partial interruptions also resulted from causes such as animal or bird contact with electrical devices, digging activity, lightning strikes, general pole damage due to rotting, termites, woodpecker holes, age, or corrosion, insulation flashover or insulation breakdown, overvoltage, undervoltage, emergency load shedding and electrical surges (which occur when de-energized lines are suddenly reenergized). Additionally, GPU identified approximately 350 pole-top transformers that experienced interruptions during the week in question, affecting approximately 3,700 customers.

According to a report by plaintiffs' technical experts, Exponent Failure Analysis Associates (Exponent), GPU recognized before July 1999 that it was operating with a deficiency in reactive power. The power outages were the result of GPU's inability to comply with its own planning criteria, and deferring upgrades for budget reasons. This deferral of projects to meet forecast demand caused low voltage, overloads and equipment failures.

Plaintiffs assert that GPU's 1998 decision to defer installation of two larger capacity transformers at the Red Bank substation triggered the worst outages. According to Exponent, the vast majority of outages beginning at 1:22 p.m. on July 6 were related to events at the Red Bank substation. That included all the planned interruptions (rolling blackouts). A GPU internal analysis stated that interruptions resulting from a fire on the number 2 transformer at the Red Bank substation, overheating on numbers 1 and 8, and automatic shutdown of number 7, affected about 100,000 customers.

The Board determined that the majority of outages were concentrated in coastal Monmouth and Ocean Counties, "primarily due to the failure of two Red Bank Substation transformer bushings and associated rolling blackouts." It determined that at Red Bank, two bushings [insulating parts] failed independently, one on July 5 and one on July 7. According to GPU, although a series of equipment failures, deliberate load reductions and protective relay trippings occurred at Red Bank within a few days, they had different causes and affected different groups of customers for varying durations.

The October 18, 1999 order certifying the class defined the class as follows:

All customers of GPU Energy in New Jersey including, but not limited to, customers in Monmouth and Ocean Counties, and all dependents, tenants, employees and other intended beneficiaries of customers of GPU Energy in New Jersey including, but not limited to, Monmouth and Ocean Counties, who suffered damages as a result of the failure of GPU Energy to deliver electricity during the week beginning Sunday, July 4, 1999. The officers, directors and agents of the defendants are specifically excluded from the class. Any persons with claims for personal injury arising out of the power outages are excluded from the class.

At a case management conference on October 18, 1999, Judge Chaiet approved the provision of a notice of claim to class members, which was inserted in customers' bills and also published in the Newark Star Ledger and the Asbury Park Press. About 2500 claims were returned, ranging in amount from $1 to $150,000. Plaintiffs do not want to rely on individual claims, claiming it is unrealistic to expect thousands of class members to testify to their relatively small damages years after the event. Accordingly, they retained two experts, engineers Roy Billinton and Garry Wacker, to estimate class-wide damages.

Those experts estimated damages at $62 million. Billinton and Wacker used existing surveys of electrical customers in California and Canada asked to value hypothetical power outages. The experts relied primarily on four cost-of-interruption studies, two Pacific Gas and Electric (PGE) studies and two of their own, in which a sample population was asked to state what their expected costs would be if an outage occurred under certain circumstances.

Billinton and Wacker adapted this data to New Jersey in order to "estimate customer damages resulting from GPU services interruptions from July 4 to 10, 1999." Billinton and Wacker explained in their report that customer surveys had "become widely accepted as the most suitable means to obtain accurate customer interruption cost data." "Typical applications of reliability worth data [estimated cost of unreliability] involve cost-benefit analyses to assist system planning decisions, justify new infrastructure, or support changes in electricity tariff structures." The actual losses due to service interruptions "are considered as a lower bound estimate or surrogate of reliability worth."

In several of their own papers on survey estimates of interruption costs, Billinton and Wacker listed some limitations of such survey results. In one paper, they cautioned that the results were predictions and estimations based on survey scenarios respondents had not experienced, and actual actions and costs might be different. In another they cautioned, "Limitations included here are: that the estimates are customer predictions of their losses or responses; that most respondents have little experience with longer interruptions; and costs of interruptions are likely socio-economic/demographic/geography specific; etc." In another they admonished:

The customer damage [cost of interruption] function is rather uncertain and incomplete for many reasons, most notable of which are the considerable variation of customer losses with each customer category and the inherent problem of combining widely varying components into a composite function. Any applications of the customer damage function are therefore similarly limited.

The authors also cautioned that "it would be necessary to generate a composite customer damage function for each region of a utility's service area or other service areas as required to undertake comparisons." "It is highly questionable, however, to make comparisons between systems or to determine the costs associated with those effective minutes of interruption." They concluded that the composite customer damage function, despite its variables and uncertainties, was "the most readily available tool for determining monetary estimates of reliability worth." "More credible values would be determined from customer cost data which are specific to the situation being studied."

Billinton knew of two studies that had used reliability worth assessment methodology to measure customer losses during an actual power interruption. A 1991 study in Norway had conducted a survey to assess the aggregate damages to customers of a blackout. An earlier Swedish study applied cost-ofinterruption data to estimate the total national cost of a 1983 blackout. Neither was done in a litigation context or even for the purpose of measuring individual damages and compensating customers.

Plaintiffs provided certifications from two other experts, population surveyor Michael J. Sullivan and engineer Douglas M. Logan. In his certification dated January 3, 2001 Sullivan said, "It is my opinion that reliable interruption cost estimates can be developed using information obtained from published literature and other sources." He explained, "Outage cost surveys are designed to accurately measure the average magnitude of the economic losses groups of customers, similar to one another in the way they use electricity, experience as a result of electric service interruptions." "Electric utilities use average interruption cost estimates (obtained from population surveys) to evaluate the benefits that can be obtained from investments in service reliability. In effect, they use avoided customer interruption costs to identify the cost effectiveness of reliability improvements."

Logan agreed with the statements of Billinton, Wacker and Sullivan about "the applicability of customer interruption costs obtained by surveys to determining the economic damages incurred by classes of customers resulting from power outages." He noted that the original purpose of those surveys was to rationalize reliability criteria and capacity pricing. Interruption costs had been used in the United States for generation planning, for tariff design and in investment decisions for transmission and distribution systems. He stated that interruption cost surveys "may be used as the basis for reliable estimates of the interruption costs for any other utility if there is a reasonable correspondence between the two utilities in terms of climate, economy, and demographics."

William Desvousges, an expert in the valuation of benefits and costs, submitted a report for defendants. He concluded that plaintiffs' methodology was improperly applied here because the value of reliability, even if assumed as accurately measured by these hypothetical studies, was not a reasonable proxy for actual damages. Moreover, the surveys used did not account for differences among customers; the customer composition and circumstances, such as season, were not comparable; and the studies had numerous methodological flaws.

Desvousges testified in deposition that it might be valid to apply data from existing surveys to another similar area, depending on the purpose of the transfer and the level of accuracy needed. It could be valid, for example, in a benefit — cost analysis or a natural resource damage assessment for purposes of settlement. Desvousges had used the benefits transfer method for the utility industry in one context only, in measuring potential environmental costs associated with utility supply plants. He conceded that the benefits transfer method might have been valid here if the surveys used by Billinton and Wacker had met the necessary criteria of soundness and similarity, but they did not. However, he qualified it by adding that such a benefits transfer method would have yielded only a "starting point," because any transfer has error associated with it.

When granting defendants' motion for decertification, Judge Perri found, "While not dispositive on the issue of decertification, the dismissal of the consumer fraud claims in this matter materially changed both the complexion of the case and the preference for certification which was present at the time Judge Chaiet decided the initial motion for certification."

Judge Perri explained that much of the law on the liberal construction of the class action rule was made in the context of consumer fraud actions, making it difficult to determine whether it was class actions, or consumer fraud actions, that actually deserved most of the credit. Moreover, the loss of the consumer fraud claims rendered plaintiffs' damages proofs more difficult. In addition, the consumer fraud claims provided a "cohesive, unifying effect" that contributed to the presence of the requirement for predominance of common issues.

Judge Perri stated that a canvass of other jurisdictions showed that, where class action status was sought for interruption of power supplies, courts had "resoundingly rejected the class action format." She found that the record showed no single catastrophic event that caused the outages, but rather, hundreds of "trouble locations" involving different equipment and distinct factual scenarios. "With the dismissal of the consumer fraud claims, plaintiffs are left with what can best be described as a mass tort action sounding in negligence and breach of contract." Such actions lack class cohesion, unless a single event is the common cause. She rejected plaintiffs' theory that "lack of preparedness" qualified for that exception, noting "[T]here is no recognized cause of action in New Jersey for `lack of preparedness'." She concluded that there was no common nucleus of proof where plaintiffs would have to prove that each separate power interruption was attributable to a failure of GPU to properly allocate its resources. Therefore, class certification was not the superior method for resolving the issues.

Even if plaintiffs could maintain class status based on their theory of liability, there remained the issue of plaintiffs' effort to prove damages on a class-wide basis. Judge Perri rejected the premise that interruption of power per se constituted loss of use, and therefore each member of the class must have sustained damage; plaintiffs would have to prove the fact of damage for each class member. For example, a shopkeeper whose store was closed, or a homeowner not in residence, might have suffered no adverse consequences despite losing power. For those plaintiffs who could prove the fact of actual damage, the nature and degree of their losses would vary greatly. The disparity among commercial customers would be even greater than that among residential customers.

Judge Perri considered the damages model proposed by plaintiffs, based on consumer surveys conducted in California and Canada on projected harm caused by hypothetical outages. She denied plaintiffs' motion to admit the testimony of plaintiffs' two experts in the field of evaluating interruption costs of electric service, Billinton and Wacker. The proposed survey method for calculating damages had been used only in the context of infrastructure planning. Customer surveys had never been used for the purpose of awarding litigation damages, and never used to provide compensation following an actual power outage; they had been used to calculate theoretical losses sustained by customers after an actual outage only in two isolated instances, in Norway and Sweden. Such survey results had not been taken from the customers of one utility company and applied to a different one.

She found the report of Desvousges persuasive, though not dispositive. The studies on which plaintiffs' experts relied differed in location, sectors, duration of outages, season, day of the week and time of day. More important, the survey method focused on the value that customers placed on the reliability of their electric service, not on actual damages possibly suffered as a result of its interruption. Also, it used averages, thus failing to take into consideration individual differences among class members.

Judge Perri observed that there is a preference for individualized proof of damages. She did not find that individualized proofs were always required, but said, "A court should only depart from this general rule where class-wide damages can be calculated by a reliable mathematical formula." She added, "Even then, a statistical model estimating the total amount of damages should not be substituted for actual proof unless it can be presumed that all members of the class suffered damage."

She concluded that the damages model proposed by plaintiffs, a prediction of losses during a hypothetical power outage, could not serve as a substitute for proof of actual damages. Judge Perri added, "No mathematical formula could accurately deduce each class member's fact-of-damage nor can it determine the actual damages sustained during outages which varied as to length and time of day." She thought the model offered by plaintiffs was "too flawed to produce an accurate representation of damages."

Judge Perri also rejected plaintiffs' proposal that the fluid recovery rule applied to relieve them of their obligation to prove actual damages. The remedy of fluid recovery is applied in a class action only when funds from an award are distributed to non-class members because the identities of class members are not known. Here, the names and addresses of GPU customers, along with the precise length of each service interruption, were readily available.

IV Standard of Review

Plaintiffs argue the judge erred by applying the standard for the initial determination of class certification instead of a more stringent standard allegedly applicable to decertification. Defendants respond that the judge acted within her discretionary authority to grant their decertification motion. We conclude that the judge applied the correct discretionary standard.

We review a determination of class certification for abuse of discretion. See In re Cadillac V8-6-4 Class Action, 93 N.J. 412, 436-37 (1983) (applying that standard). Class certification is governed by R. 4:32-1, which provides:

(a) General Prerequisites to a Class Action. One or more members of a class may sue or be sued as representative parties on behalf of all only if (1) the class is so numerous that joinder of all members is impracticable, (2) there are questions of law or fact common to the class, (3) the claims or defenses of the representative

parties are typical of the claims or defenses of the class, and (4) the representative parties will fairly and adequately protect the interests of the class.

(b) Class Actions Maintainable. An action may be maintained as a class action if the prerequisites of paragraph (a) are satisfied, and in addition:

(1) the prosecution of separate actions by or against individual members of the class would create a risk either of (A) inconsistent or varying adjudications with respect to individual members of the class which would establish incompatible standards of conduct for the party opposing the class, or (B) adjudications with respect to individual members of the class which would as a practical matter be dispositive of the interests of the other members not parties to the adjudications or substantially impair or impede their ability to protect their interests; or

(2) the party opposing the class has acted or refused to act on grounds generally applicable to the class, thereby making appropriate final injunctive relief or corresponding declaratory relief with respect to the class as a whole; or (3) the court finds that the questions of law or fact common to the members of the class predominate over any questions affecting only individual members, and that a class action is superior to other available methods for the fair and efficient adjudication of the controversy. The factors pertinent to the findings include: first, the interest of members of the class in individually controlling the prosecution or defense of separate actions; second, the extent and nature of any litigation concerning the controversy already commenced by or against members of the class; third, the difficulties likely to be encountered in the management of a class action.

Class actions "should be liberally allowed where consumers are attempting to redress a common grievance under circumstances that would make individual actions uneconomical to pursue." Varacallo v. Mass. Mut. Life Ins. Co., 332 N.J. Super. 31, 45 (App.Div. 2000). An applicant must establish the threshold requirements of numerosity, commonality, typicality and adequacy of representation set forth in subpart (a) of R. 4:32-1, and also satisfy one of the three alternative requirements set forth in subpart (b). Saldana v. City of Camden, 252 N.J. Super. 188, 193 (App.Div. 1991). Under R. 4:32-1(b)(3), the party seeking certification must demonstrate both predominance of the common issues and superiority of a class action over other trial techniques. Id. at 196.

When granting class certification here, Judge Chaiet found that the R. 4:32-1(a) requirements of numerosity, commonality, typicality and adequacy of representation were satisfied. He also found that plaintiffs satisfied the requirements of R. 4:32-1(b)(3). He found a predominance of common, over individual, questions of law and fact, because, even though there were claims of fraud, the surrounding issue was defendants' negligence in failing to supply electricity despite knowledge that peak demand periods would occur. He added that individual actions would be expensive.

Rule 4:32-2 controls the determination of maintainability of class actions, notice, judgments, and partial class actions. The part of that rule pertinent to decertification is R. 4:32-2(a), which provides:

Order Determining Maintainability. As soon as practicable after the commencement of an action brought as a class action, the court shall determine by order whether it is to be so maintained. An order under this subdivision may be conditioned, and may be altered or amended before the decision on the merits.

Judge Chaiet expressly anticipated the possibility of decertification when he granted certification, saying:

[T]he Court is not married to the concept of a class action in this case. And I will closely monitor it to see that it is accomplishing the appropriate goal of the class action suit to minimize expenses and maximize participation to fairly resolve a widespread problem. The Court will not hesitate to curtail this action under Rule 4:32-2(a) if the goals are not being met.

New Jersey courts have recognized that, under R. 4:32, in an appropriate case, a trial court can decertify a class. In re Cadillac, 93 N.J. at 437. Class certification decisions rest in the sound discretion of the trial court. See id. at 436 (class action certification reviewed for an abuse of discretion). However, the parties do not dispute that, as Judge Perri noted, the standard applied on a motion to decertify a class has never been directly addressed in New Jersey.

"Our class-action rule, R. 4:32, is a replica of Rule 23 of the Federal Rules of Civil Procedure as amended in 1966." Riley v. New Rapids Carpet Ctr., 61 N.J. 218, 228 (1972). Construction of the federal rule may be considered helpful, if not persuasive, authority. See Saldana v. City of Camden, 252 N.J. Super. at 194 n. 1. Rule 4:32-2 tracks Rule 23(c) of theFederal Rules of Civil Procedure. Fed.R.Civ.P. 23(c)(1) provides: "As soon as practicable after the commencement of an action brought as a class action, the court shall determine by order whether it is to be so maintained. An order under this subdivision may be conditional, and may be altered or amended before the decision on the merits." The content of the federal rule matches that of R. 4:32-2(a) except for New Jersey's use of "conditioned" instead of "conditional" and the addition of the title "Order Determining Maintainability."

A party seeking class certification has the burden of proof. Davis v. Romney, 490 F.2d 1360, 1366 (3d Cir. 1974). The burden of persuasion remains with the party which desires to maintain certification. Smith v. Armstrong, 968 F. Supp. 50 , 53 (D. Conn. 1997).

After a judge enters a certification order, "the judge remains free to modify it in the light of subsequent developments in the litigation," because conformance to the class action rule is indispensable. General Tel. Co. of the Southwest v. Falcon, 457 U.S. 147, 160, 102 S. Ct. 2364, 2372, 72 L. Ed. 2d 740, 752 (1982). A certification order is "inherently tentative," particularly before notice is sent to class members. Coopers Lybrand v. Livesay, 437 U.S. 463, 469 n. 11, 98 S. Ct. 2454, 2458 n. 11, 57 L. Ed. 2d 351, 358 n. 11 (1978). Plaintiffs do point out that numerous cases from other jurisdictions place on a party seeking to decertify a class a heavy burden to produce evidence of changed circumstances or new facts or law.

For example, in Slaven v. BP America, Inc., 190 F.R.D. 649 , 651 (C.D. Cal. 2000), the court stated that because a party seeking class certification bears the burden of demonstrating that the proposed class satisfies the elements of the class action rule, a party seeking decertification should bear the burden of demonstrating that it does not. The Slaven court characterized the defendants' burden in urging decertification as "relatively heavy, since `doubts regarding the propriety of class certification should be resolved in favor of certification.'" Ibid. (quoting Groover v. Michelin North Amer., Inc., 187 F.R.D. 662 , 670 (M.D. Ala. 1999)). However, that court also cautioned that a motion for decertification should not be treated like a motion for reconsideration; that is, it should not be entertained only reluctantly.Id. at 651-52. The court emphasized that the federal class action rule itself provided for periodic reassessment of class rulings. Id. at 652.

Similarly, in Gordon v. Hunt, 117 F.R.D. 58 , 61 (S.D.N.Y. 1987), the court concluded that "defendants bear a heavy burden to prove the necessity of either the drastic step of decertification or the less draconian but still serious step of limiting the scope of the class." InDoe v. Karadzic, 192 F.R.D. 133 , 137 n. 6 (S.D.N.Y. 2000), the court distinguished Gordon v. Hunt because in Karadzic, a group of dissatisfied plaintiffs, not the defendants, sought decertification; moreover, unlike in Gordon v. Hunt, no massive effort had been undertaken to identify and notify potential plaintiffs. Karadzic nevertheless held that a court may not disturb its prior class certification absent "`some significant intervening event'" or "`a showing of compelling reasons to reexamine the question.'" Id. at 136-37 (quoting Langley v. Coughlin, 715 F. Supp. 522 , 533 (S.D.N.Y. 1989), appeal dismissed, 888 F.2d 252 (2d Cir. 1989), and Wilder v. Bernstein, 645 F. Supp. 1292 , 1311-12 (S.D.N.Y.), aff'd, 848 F.2d 1338 (2d Cir. 1988)).

In contrast, in Ellis v. Elgin Riverboat Resort, 217 F.R.D. 415 , 419 (N.D. Ill. 2003), the court concluded that the party seeking certification bore the burden of producing evidence demonstrating that the class action elements were satisfied on a motion to decertify as well as on the initial motion for certification. Accord Chisolm v. TranSouth Financial Corp., 194 F.R.D. 538 , 544, 548 n. 5 (E.D. Va. 2000) (same legal standard for decertification motion as for class certification motion; disagreeing with Langley v. Coughlin, 715 F. Supp. at 552 ). Though the Ellis court reached a conclusion contrary to that of theSlaven court concerning the burden of producing evidence, like theSlaven court it relied on the provision of the class action rule assigning to the court the continuing obligation to review the propriety of a class action, and the "inherently tentative" nature of class certification. Ellis v. Elgin Riverboat Resort, 217 F.R.D. at 419 (quoting Coopers Lybrand v. Livesay, 437 U.S. at 469 n. 11,98 S. Ct. at 2458 n. 11, 57 L. Ed. 2d at 358 n. 11).

Here, the judge did not expressly place on either party the burden of producing evidence, but invoked the principle of Lusky v. Capasso Bros., 118 N.J. Super. 369, 373 (App.Div.), certif. denied, 60 N.J. 466 (1972), that "the class action rule should be liberally construed, and such an action should be permitted unless there is a clear showing that it is inappropriate or improper." In effect, the judge placed on defendants the burden of producing evidence that class treatment was inappropriate. The judge here said, "If, upon considering the record in accordance with the liberal construction required under Lusky, the court finds that the criteria for and goals of class certification are no longer being met, the motion for decertification should be granted." Plaintiffs do not convince us that this standard was incorrect.

V Reliance on Dismissal of CFA claims

Plaintiffs next argue that the judge erred by finding the dismissal of the CFA claims had changed the complexion of the case and removed a preference for class certification. To support her view that dismissal of the CFA claims was significant, Judge Perri cited Carroll v. Cellco Partnership, 313 N.J. Super. 488, 498 (App.Div. 1998). There we pointed out that, in New Jersey, class actions are liberally construed and permitted absent a clear showing of impropriety; we added, "This preference for class certification applies especially for adjudication of multiple consumer-fraud claims." Ibid.

Judge Perri correctly observed that proof of damage in CFA cases is easier than proof of damage in negligence cases. To establish a cause of action for consumer fraud, a plaintiff need show only that a defendant engaged in a practice unlawful under the CFA that resulted in an "ascertainable loss of moneys or property, real or personal." N.J.S.A. 56:8-19. In contrast, to establish a cause of action for negligence, a plaintiff must prove breach of a duty of care and actual damages sustained as a proximate cause of the breach. Weinberg v. Dinger, 106 N.J. 469, 484 (1987).

Plaintiffs assert that Judge Chaiet did not expressly rely on the CFA claims, but granted certification based on "negligence issues." We agree. Judge Chaiet stated:

The Court is satisfied that the common issues of fact and law do exist. This is primarily a case in which there is a claim that GPU through negligence failed to supply the needs of its customers. In Red Bank the issue concerns the two generators that failed. In the other towns the issue concerns the transformers, but the overriding issue is GPU's lack of preparedness to meet the demand

However, the absence of express reliance thereon does not negate the significance of the CFA claims.

Plaintiffs also contend that Judge Chaiet's citations to CFA cases did not mean that CFA claims were necessary to maintain class certification. Again, we agree. As plaintiffs point out, the principle that the class action rule should be liberally construed has been applied in cases without statutory consumer fraud claims. E.g., Saldana v. City of Camden, 252 N.J. Super. at 196 (accord in negligence action); Lusky v. Capasso Bros., 118 N.J. Super. at 373 (stating principle in contract and tort action). However, there is an enhanced preference for class actions in consumer fraud cases. See Strawn v. Canuso, 140 N.J. 43, 68 (1995) ("[A] class action is the superior method for adjudication of consumer-fraud claims."). But Judge Perri did not conclude that the loss of the CFA claims was dispositive per se on the issue of decertification.

Plaintiffs also contend that, assuming there is a special preference for class action status accorded consumer fraud cases, certification is warranted here. That is because this case, even stripped of its statutory CFA cause of action, involves allegations of consumer fraud. The term "consumer fraud" is ordinarily used to refer to an action under the CFA, in contrast to common law fraud. E.g., Varacallo v. Mass. Mut. Life Ins. Co., 332 N.J. Super. at 45 (comparing and contrasting the two). Nevertheless, in In re Cadillac, 93 N.J. at 435, the Court said it was "mindful that the class action rule should be construed liberally in a case involving allegations of consumer fraud," without specifying a statutory cause of action.

The Cadillac Court affirmed the certification of a statewide class of purchasers of automobiles with a particular type of engine, where the causes of action included breach of express and implied warranties, negligence, strict liability, fraud, misrepresentation, violation of the federal Magnuson-Moss Warranty Act, and refusal to accept rescission. In re Cadillac, 93 N.J. at 423, 426-30. The Supreme Court stated, "Certification of a class action should not be denied because of the underlying theory on which the action is predicated." Id. at 426.

None of plaintiffs' arguments convince us that Judge Perri erred in her finding that dismissal of the CFA claims were significant. The Supreme Court, while cautioning that certification of a class action should not be denied based on the underlying legal theory, also recognized that "even the identification of the issues to determine the suitability of an action for certification requires some preliminary analysis." In re Cadillac, 93 N.J. at 426. As we have noted, Judge Perri expressly found not that the loss of a legal theory was dispositive, but that its loss weighed against a class action preference and altered plaintiffs' burden to prove damages.

In Riley v. New Rapids Carpet Center, 61 N.J. at 228, a consumer fraud case, the Court cautioned that "a court should be slow to hold that a suit may not proceed as a class action." Although R. 4:32-2(a) mandates the determination of the maintenance of a class action "as soon as practicable" after commencement of the action, especially in a CFA case, plaintiffs may have difficulty early in the proceedings demonstrating that the elements are satisfied. Ibid. Here, however, there has been no problem posed of a hasty rejection of the class.

VI Lack of Predominance

Plaintiffs next argue that the judge erred by finding a lack of predominance of common issues. Defendants respond that the court correctly found no single event that would have caused the alleged harm and correctly found that individual issues predominated on causation and on proof of injury. The need for individual damages calculations was not alone dispositive. We conclude that decertification was justified by the predominance of individual causation issues. However, certification is appropriate for a more limited class of customers affected by the failures of equipment at the Red Bank substation.

Judge Perri did not expressly find that plaintiffs had satisfied the threshold requirements of numerosity, commonality, typicality and adequacy of representation set forth in R. 4:32-1(a), but we agree with plaintiffs that she impliedly so found. Judge Chaiet had so found. We also agree that Judge Perri concluded that plaintiffs failed to satisfy only the requirement of R. 4:32-1(b)(3), for predominance of common questions of law or fact and the superiority of a class action. The predominance requirement is more demanding than the commonality requirement. Amchem Prods., Inc. v. Windsor, 521 U.S. 591, 623-24, 117 S. Ct. 2231, 2250,138 L. Ed. 2d 689, 713 (1997) (asbestos claims). Plaintiffs do not argue that they are entitled to maintain the action as a class action because, in the alternative, they satisfied the requirements of either subpart (1) or (2) of R. 4:32-1(b). Absent those alternatives, failure to satisfy the predominance requirement is fatal.

Plaintiffs contend that Judge Perri, viewing the case as a mass tort, wrongly concluded that they failed to prove a single event that caused their harm. They identify the single event as "the failure of GPU's interconnected delivery system to meet the foreseen demand triggered by the weather." Plaintiffs also contend that, even without a single event, it was sufficient that the outages had a common cause, which was "GPU's decision to cut projects necessary to meet anticipated load." They find defendants' focus on the direct causes of outages to particular customers, such as equipment failures, too narrow.

Judge Perri found that "there is no recognized cause of action in New Jersey for `lack of preparedness'." But plaintiffs assert that such a theory was recognized in Weinberg v. Dinger, 106 N.J. 469. We disagree. In Weinberg, the Court abrogated the longstanding immunity of water companies for losses caused by negligent failure to maintain adequate water pressure for fire fighting, except with respect to subrogation claims from insurers. Id. at 492. That did not commit the Court to the recognition of a theory that "lack of preparedness" constituted a single event as the alleged cause of a mass tort.

Here, the judge found, "With the dismissal of the consumer fraud claims, plaintiffs are left with what can best be described as a mass tort action sounding in negligence and breach of contract." The judge relied on Amchem Products, Inc. v. Windsor, 521 U.S. at 625,117 S. Ct. at 2250, 138 L. Ed. 2d at 713-14, for the proposition that mass accident cases are ordinarily not appropriate for class treatment because they "are likely to present `significant questions, not only of damages but of liability and defenses of liability, . . . affecting the individuals in different ways.'"

In the portion of Amchem quoted by Judge Perri, the Supreme Court was quoting the Advisory Committee for the 1966 Revision of Rule 23. However, the Amchem Court prefaced the quotation from the Committee: "Even mass tort cases arising from a common cause or disaster may, depending upon the circumstances, satisfy the predominance requirement."Ibid., 117 S. Ct. at 2250 , 138 L. Ed. 2d at 713. It cautioned that, despite that 1966 comment, the text of Rule 23 did not "categorically exclude mass tort cases from class certification, and District Courts, since the late 1970's, have been certifying such cases in increasing number." Ibid., 117 S. Ct. at 2250 , 138 L. Ed. 2d at 714.

Plaintiffs complain that by omitting these statements, the court "dramatically misquot[ed]" the Supreme Court case. We disagree. The judge did not misquote or even mischaracterize, though she did not set forth the entire reasoning of the Supreme Court. Plaintiffs themselves omit a subsequent sentence concerning the trend toward certifying class actions in mass tort cases: "The Committee's warning, however, continues to call for caution when individual stakes are high and disparities among class members great." Ibid., 117 S. Ct. at 2250 , 138 L. Ed. 2d at 714. InAmchem, the Court affirmed the decertification of a class for failure to satisfy the predominance requirement of Fed.R.Civ.P. 23(b)(3). The judge's reliance on Amchem was not misplaced.

Plaintiffs distinguish Saldana, 252 N.J. Super. 188, on which Judge Perri also relied. In Saldana, owners of properties damaged by fires started in abandoned buildings owned by the City of Camden sued the city and certain city officials and sought class certification. Id. at 190. The plaintiffs' common theory was that the defendants had failed to implement or administer a policy to adequately regulate the city-owned buildings. Id. at 193. Judge Perri analogized to the theory of negligence alleged by plaintiffs here.

However, in Saldana, the court found the issues of liability, causation and damages would require individualized treatment, because "over 80 privately-owned dwellings were damaged by 61 separate fires over a six-year period," and proof of the causes and damages would be fact-specific, including the possible presence of intervening causes.Id. at 197. Judge Perri found a similar variation in the proofs required here, because plaintiffs must prove that each power interruption resulted from GPU's failure to properly allocate its resources.

Plaintiffs distinguish Saldana because GPU had sole control of its system, and the outages occurred "during a common weather event over a single period of a few days." However, the clear evidence showed that there were multiple causes for the power interruptions in various areas. GPU could be held liable for all of them only if the court accepted plaintiffs' theory that GPU's policy leading to a lack of preparedness itself constituted negligence, and that negligence was the proximate cause of all outages.

Plaintiffs rely on Fox v. Cheminova, Inc., 213 F.R.D. 113 (E.D.N.Y. 2003), in which the court certified a class of lobstermen alleging that use of pesticides in September and October 1999 to control mosquitoes, carried into Long Island Sound by runoff hastened by a tropical storm, had caused lobster mortality. There the defendants argued that certification was inappropriate in a mass tort case, relying on Amchem, 521 U.S. 591, 117 S. Ct. 2231, 138 L. Ed. 2d 689. The court distinguishedAmchem factually and found predominance, agreeing with the plaintiffs that the alleged damage had occurred within a short time frame; all potential class members suffered losses resulting from the die-off; and the causal mechanism, pesticide exposure, was the same for all. 213F.R.D. at 129. The court found the die-off analogous to a single catastrophic event, like an airplane crash, and noted that specific issues of causation should be left for trial. Ibid.

Plaintiffs contend that similarly here, common liability issues predominate because the breach of contract claims are based on a common tariff and negligence claims are based on the common allegation that all outages were caused by GPU's systemwide failure to meet the increased forecast demand Plaintiffs focus on defendants' duty to maintain the power supply. This does not answer Judge Perri's concern that there was not one outage, but many, whose direct causes were admittedly numerous. Plaintiffs' theory of common liability rests on their premise that a general lack of preparedness itself constituted negligence and breach of contract, a premise that the judge correctly rejected.

Plaintiffs also contend Judge Perri wrongly concluded that individual damages issues would predominate over common ones. She noted that the nature and degree of the losses of proposed class members would vary greatly. We agree with plaintiffs that that this likely variation would not necessarily preclude certification. Plaintiffs point to Delgozzo v. Kenny, 266 N.J. Super. 169 , 190 (App.Div. 1993), in which the court said that, if common questions of liability predominated, a need for individual damages calculations would not preclude class action certification. See also Central Power Light Co. v. City of San Juan, 962 S.W.2d 602, 610 (Tex.App. 1998) ("[C]lass certification will not be prevented merely because damages must be determined separately for each member of the class.") Here, however, the court reasonably found that common questions of liability did not predominate.

The court found that, among residential customers, damages varied not only in degree, depending on the length of interruption of power, but also in kind, depending on individual circumstances; among commercial customers the disparities might be even greater. Plaintiffs proposed to overcome the problem of individualized proof of damages by offering class-wide proof, but the court rejected that approach, as we will discuss.

The parties discuss several cases from other jurisdictions where utilities customers attempted to bring class actions against a provider. When noting that other jurisdictions had resoundingly rejected the class action format for suits based on interruption of power supplies, Judge Perri cited three Louisiana cases and one Texas case. In Brown v. New Orleans Public Service, Inc., 506 So.2d 621 (La.Ct.App.), writ denied, 508 So.2d 67 (La. 1987), utility customers sued for power interruptions during a cold snap. The Louisiana appellate court reversed class certification, explaining that, although the alleged negligence was common to all claims, there were too many individual differences in proof of causation and the availability of defenses like comparative negligence, as well as the amount of damages, to warrant a finding of predominance of common facts and issues. Id. at 623.

In LaFleur v. Entergy, Inc., 737 So.2d 761 (La.Ct.App. 1998), another winter storm power interruption case where the plaintiffs alleged negligent failure to maintain the power system, the appellate court affirmed the trial court determination that the proposed class did not meet the predominance requirement. Although an ice storm ultimately caused power outages affecting all plaintiffs, it was not clear that the damages of the entire putative class were caused by the defendants' alleged negligence. Id. at 765.

In Entergy Gulf States, Inc. v. Butler, 25 S.W.3d 359 (Tex.App. 2000), the Texas appellate court reversed a class certification granted to electrical utility customers who claimed damages from power outages triggered by an ice storm. Again the plaintiffs based their claim on the alleged failure to properly maintain the system to prepare for storms, and again the court found that common issues did not predominate. Id. at 362-63. The court found in addition that the inclusion of claims for personal injuries weighed against class certification. Id. at 363.

In Royal Street Grocery, Inc. v. Entergy New Orleans, Inc., 778 So.2d 679 (La.Ct.App.), writ denied, 789 So.2d 594 (La. 2001), the appellate court affirmed the trial court's denial of class certification where a fire in an underground electrical vault prompted the fire department to twice order that the utility de-energize an entire grid. Business and individual electrical customers sued, and the court found a lack of predominance of common liability issues. Id. at 685. Various parties had different causes of action and even different standing issues, and the individualized claims could warrant different defenses. Ibid.

Plaintiffs contend that these cases were either distinguishable on the facts, took too cautious an approach to certification, or were wrongly decided. On the contrary, we find these cases are quite similar in structure to the instant case, and, though from sister states, support Judge Perri's reasoning.

Another case that supports defendants' position is Tegnazian v. Consolidated Edison, Inc., 730 N.Y.S.2d 183 (Sup.Ct. 2000). There, too, customers of an electric utility who had lost power in a blackout sought damages, and the court denied class certification based on the absence of the predominance of common questions of law and fact. The court noted that the major common inquiry was whether the utility was grossly negligent in failing to provide service, but that in itself was not sufficient to justify class certification. Id. at 186-87. Other issues, including standing, liability, causation and the existence and amount of damages, would require individual inquiry and would predominate. Id. at 186.

Plaintiffs contend that the judge here wrongly failed to mention several cases that supported their arguments. But any support those cases provide is weak at best. In one, Bucci v. Cunard Line Ltd., 35 Pa. D. C. 3d 228 (Ct. Common Pleas 1985), the court granted class certification to approximately 400 cruise ship passengers alleging damages resulting from a power failure on the ship. The court found that common questions of law and fact predominated because the primary issue was breach of the cruise line's duty to provide the essential services for which the passengers contracted, and every passenger suffered a disruption of those services. Id. at 232-35. This case is quite distinguishable on the facts. Power failure on a cruise ship was likely attributable to a single cause, and the passengers' damages would have been relatively uniform since all were in similar circumstances and subject to the same privations at the same time. Courts in other jurisdictions have rejected class actions in the context of passenger cruises. See Stobaugh v. Norwegian Cruise Line, Ltd., 105 S.W.3d 302, 311 n. 4 (Tex.App. 2003) (distinguishing Bucci based on Texas's more "rigorous analysis" of class action applications and listing cases with contrary results).

In a second case cited by plaintiffs, McDonald v. Washington, 862 P.2d 1150 (Mont. 1993), water company customers sued the supplier for negligent maintenance of the system, alleging that tap water was neither clean nor drinkable, and the trial court certified the class. The Montana Supreme Court affirmed the certification, noting when finding the predominance of common facts and issues that all counts of the complaint stemmed from the alleged failure of the company to provide potable water and adequate service. Id. at 1155. The court distinguished Brown v. New Orleans Public Service, 506 So.2d at 623, because that case presented issues of comparative negligence and intervening cause.

Plaintiffs further argue that, at Red Bank, there was a single causative event, transformer failures, which would have permitted a Red Bank class, at least. Defendants respond that the Red Bank outages did not constitute a single event, nor could they be attributed to any negligence on the part of GPU, and that customers who lost power as a result of Red Bank equipment failures would have to show that negligence caused the particular failure that led to the relevant outage.

A Board report found that GPU had deferred to the year 2000 the installation of new transformers at Red Bank, originally planned for 1998. It found that the management decision to defer the installation was "risky," and partly motivated by manpower and budgetary constraints. Plaintiffs' theory was that the deferral constituted negligence, and caused the bushing failures in 1999. However, the validity of the negligence claim is technically not before the court on this appeal; the relevant question is whether the installation deferral might have caused all the outages attributable to the Red Bank substation.

During the decertification hearing, the court questioned plaintiffs' counsel concerning the possibility of limiting the class to those affected by the transformer failures at the Red Bank substation. Plaintiffs' counsel described Red Bank as a microcosm of the larger problem, demonstrating the same domino effect but with events more compactly interrelated than those overall. Plaintiffs' counsel asserted that GPU's own documents showed that 74% of circuit outages were related, or possibly related, to equipment failures in Red Bank. Defense counsel asserted that, even within that one substation, there were a series of events that led to power interruption, many of which were unrelated to each other. Judge Perri did not address the merits of certifying a reduced, Red Bank, class.

Exponent, plaintiffs' expert, accepted a summary of the Red Bank equipment failures prepared by Stone Webster Management Consultants, Inc., a consultant retained by the Board. That report pointed to a bushing failure on transformer bank number 2 on July 5, followed by explosion and fire, and another bushing failure on transformer bank number 1 on July 6 and 7. The report continued, "The loss of the second transformer caused subsequent overloads on the remaining transformers (numbers 7 and 8) which resulted in interruption of service to approximately 105,000 customers."

Defendants contend that there is no connection between the deferred installation of new transformers and the bushing failures, because the installation was not planned in order to prevent bushing failures, and there was no evidence that new transformers would not have experienced bushing failures. This ignores the Board report findings that the bushing failures in transformer bank number 1 and transformer bank number 2 resulted primarily from long-term insulation degradation, exacerbated by elevated temperatures and overvoltages. New transformers would have included new bushings.

Defendants also contend that their initial substation reinforcement project would have added a fifth transformer, leaving numbers 1 and 2 (and 7 and 8) in place, and thus permitting the failures that actually occurred. However, according to Exponent, though a 1996 study proposed a fifth transformer, a later study proposed replacement of transformer banks 1 and 2. Moreover, presumably an additional transformer would have reduced the load on the others, reducing the probability of failure through overload.

We agree with plaintiffs that a Red Bank class would be proper if their evidence could establish that failure to timely install two new transformers constituted negligence and that this failure was the proximate cause of the related outages. The lack of predominance in causation issues for the larger proposed class is much reduced in a Red Bank class. A court must not make a preliminary decision on the merits when determining whether a class should be certified. Delgozzo v. Kenny, 266 N.J. Super. at 180-81. However, the court is bound to take the substantive allegations of the complaint as true. Id. at 181.

Once the court finds that common questions of liability, and the fact of damage, predominate, individual variations in the calculation of damages does not preclude class certification. Id. at 190. However, plaintiffs may be required to present individualized proof of damages. The judge did not err by finding a lack of predominance of common liability issues for the overall class. However, we remand for certification of a limited, Red Bank, class.

VII Proof of Damages

Plaintiffs argue that the court erred by rejecting their proposed means of proving class-wide damages. We disagree.

Judge Perri did not find that individualized proofs were always required, but said that a court should depart from the general rule preferring them only where class-wide damages could be "calculated by a reliable mathematical formula." She found that reliance on a statistical model required a presumption that all members of the class had suffered damage.

The judge rejected a presumption here that every class member must have sustained damage because interruption of power per se constituted loss of use. She agreed with plaintiffs that damages could be recovered based on loss of use alone, without the need to prove pecuniary loss. However, she reasoned that the fact of damage must still be proved, because loss of use is compensable only when it causes actual harm.

The judge concluded that the damages model proposed by plaintiffs could not substitute for proof of actual damages. She found that no mathematical formula was appropriate here, because there was no valid presumption of damage to every class member. In addition, the model offered here was "too flawed to produce an accurate representation of damages."

a. Permissibility of class-wide proof of damage

Plaintiffs contend that, although the issue of class-wide aggregate proof of damages has not been addressed in New Jersey, numerous courts in other jurisdictions have endorsed it. Plaintiffs deny that they sought a "presumption" of damages for everyone who lost power, contending that they offered proof that everyone was damaged to some extent. Plaintiffs explain that the evidence actually showed, and defense witnesses admitted, that everyone who lost power suffered at least some inconvenience, if only having to reset clocks.

Here the judge did not find that class-wide proof of damages was never permitted. As defendants recognize, she found instead that a court should depart from the general preference for individualized proof only where a reliable mathematical formula is available and the court can presume that all class members suffered damage.

If aggregate damages were permitted in this case, their calculation would have to be based on estimates of individual damages, requiring the collection of data including the actual time, place and duration of outages and the customers' ordinary usage. In In re Prudential Insurance Co. of America Sales Practices Litigation, 962 F. Supp. 450 , 517 n. 46 (D.N.J. 1997), aff'd in part, vacated in part on other grounds, 148 F.3d 283 (3d Cir. 1998), cert. denied, 525 U.S. 1114,119 S. Ct. 890, 142 L. Ed. 2d 789 (1999), the court stated that when there is sufficient evidence of the fact of damage, an intelligent estimate of the amount is acceptable. It approved a settlement for a class action against a life insurer for deceptive sales practices. Id. at 468. The court acknowledged that the possible methods of proving the amount of damages included expert testimony estimating aggregate damages sustained by the class, or a formula to be applied to individual class members to compute damages during the claims allocation process. Id. at 517. It noted, however, that "sufficient facts must be introduced so that a court can arrive at an intelligent estimate without speculation or conjecture."Id. at 517 n. 46 (quoting Rochez Bros., Inc. v. Rhoades, 527 F.2d 891, 895 (3d Cir. 1975), cert. denied, 425 U.S. 993, 96 S. Ct. 2205,48 L. Ed. 2d 817 (1976)). See In re Sugar Indus. Antitrust Litig., 73 F.R.D. 322 , 353 (E.D. Pa. 1976) (damages need not be established with absolute certainty, but may not be determined by mere speculation or guess).

Plaintiffs rely on Camaraza v. Bellavia Buick Corp., 216 N.J. Super. 263, 267 (App.Div. 1987), for the proposition that inconveniences caused by loss of use of an automobile, where that loss of use was attributed to wrongful conduct of a tortfeasor, were compensable. That is, it was not necessary to incur the expense of renting a substitute vehicle to prove damages. Ibid. We find Judge Perri did not misapply Camaraza, but correctly recognized its holding that economic outlay was not necessary to prove harm did not amount to a holding that harm could always be inferred from loss of use alone. For example, a residential customer not in residence during a power loss, or a commercial customer whose store was closed, might have no damages except the inconvenience of resetting clocks.

Plaintiffs further contend that even with neither presumption nor proof of class-wide damages, aggregate damages were permitted. They find error in Judge Perri's assertion that a class-wide model cannot be used unless all members of the class suffered damage, contending that aggregate proofs are permitted as long as most members suffered damage. The court relied on Newton v. Merrill Lynch, Pierce, Fenner Smith, Inc., 259 F.3d 154 (3d Cir. 2001), a securities case. There, the court distinguished between the fact of damage and the quantum of damage, saying: "While obstacles to calculating damages may not preclude class certification, the putative class must first demonstrate economic loss on a common basis. As noted, the issue is not the calculation of damages but whether or not class members have any claims at all." Id. at 189.

The court in Newton reversed the district court's determination of class certification because not all class members were injured by the alleged securities violation — the failure of the brokers to find the best possible price. Id. at 187-88. The court explained: "[O]nly those class members whose trades could have been executed at better prices sustained economic injury here. Determining which class members were economically harmed would require an individual analysis into each trade and its alternatives. The individual questions, therefore, are overpowering." Id. at 188-89.

Plaintiffs distinguish Newton, explaining that there, the named plaintiffs rarely suffered damage in the securities trades at issue, whereas here, all the named plaintiffs showed that they sustained out-of-pocket damages, implying that other class members did, too. This might support an argument for a presumption of class-wide harm, but it does not show why the reasoning of Newton was not applicable here. The district court in Newton concluded that where an undefined number of class members sustained no economic loss, the plaintiffs' claims would require individual treatment to determine actual injury, and therefore damages were not susceptible to class-wide proof; the Third Circuit agreed. Id. at 178-79.

Plaintiffs point to In re Linerboard Antitrust Litigation, 305 F.3d 145, 155-58 (3d Cir. 2002), cert. denied, 538 U.S. 977, 123 S. Ct. 1786,155 L. Ed. 2d 666 (2003), an antitrust action that factually distinguishedNewton. Plaintiffs assert that in Linerboard, the court "approved aggregate damage proofs comparable to those here even though some class members had no damages." Plaintiffs also say that together, Linerboard and Newton "teach that aggregate proof of damages is available where the vast majority of class members sustained damages, but not where most class members did not." Both characterizations are contradicted by theLinerboard opinion itself. We agree with defendants that in Linerboard, the Third Circuit found that aggregate proof of damages was permitted because there was a fair presumption of economic loss to every class member.

The Linerboard court expressly said that the district court did not err in determining that the plaintiffs could establish injury on a class-wide basis because the proffered experts showed they could "authenticate their professional opinions that all class members would incur such damages," and therefore the plaintiffs could "prove antitrust impact common to all members of the class." Id. at 155 (emphasis added). Contrasting Newton, the Linerboard court said, "First, in Newton it was clear that not all members of the putative class sustained injuries; here, all members sustained injuries because of the artificially increased prices." Id. at 157.

Plaintiffs cite two New Jersey cases to support their proposition that class certification may be proper although not all class members have sustained damage, Varacallo v. Mass. Mut. Life Ins. Co., 332 N.J. Super. at 48, and Lusky v. Capasso Bros., 118 N.J. Super. at 373. In Varacallo, this court found that a class was permitted although there was a small possibility that there was a superseding cause for losses suffered by a few policyholders. In Lusky, this court found that dismissal of a class action by property owners against a garbage removal service was not warranted merely because there might be some individual agreements that were not breached by the defendants. 118 N.J. Super. at 373. Neither case addresses this issue, which was not maintenance of the class despite the absence of damages to every class member, but the use of classwide proof of damages in those circumstances.

Plaintiffs point to five cases from other jurisdictions in which a court accepted aggregate damage evidence. All required proof, or a presumption, that all class members were damaged, and a reliable method of allocating damages among individual class members. The cases cited by plaintiffs support the judge's conclusion that a court should depart from the general preference for individualized proof only where a reliable mathematical formula is available and the court can presume that all class members suffered damage.

In Avery v. State Farm Mutual Automobile Insurance Co., 746 N.E.2d 1242, 1260 (Ill.App.Ct. 2001), appeal allowed, 786 N.E.2d 180 (Ill. 2002) (no further history available), the court permitted damages for installation of a replacement part and loss of use of an automobile (labor and rental costs) to be measured based on extrapolation to all eligible class members from a representative sample. In In re Prudential Ins. Co. of Amer. Sales Practices Litig., 962 F. Supp. at 517 , in the context of a settlement, the court approved the use of expert testimony to estimate the aggregate amount of damages to life insurance policy holders for deceptive sales practices. In Iron Workers Local Union No. 17 Insurance Fund v. Philip Morris Inc., 29 F. Supp. 2d 825, 838 (N.D. Ohio 1998), the court recognized that "a formula to compute class-wide damages is an accepted approach if sufficiently reliable to justly determine the defendants' liability," though it did not actually approve the damage model offered by the plaintiffs.

In Catlett v. Missouri Highway Transportation Commission, 828 F.2d 1260, 1267 (8th Cir. 1987), cert. denied, 485 U.S. 1021,108 S. Ct. 1574, 99 L. Ed. 2d 889 (1988), a gender discrimination case, the court found it acceptable to award class-wide relief, in the form of back-pay, and divided pro rata among qualified class members, where identification of the individuals entitled to relief was not possible. The court noted that all class members were presumptively victims of discrimination absent proof to the contrary, which was not presented.

In Boeing Co. v. Van Gemert, 444 U.S. 472, 475-76, 100 S. Ct. 745, 748, 62 L. Ed. 2d 676, 680 (1980), the lower court awarded aggregate damages for a securities law violation. The plaintiffs had received inadequate notice of a deadline for the exercise of conversion rights of certain debentures. Id. at 474, 100 S. Ct. at 748, 62 L. Ed. 2d at 679. Damages were calculated as the difference between the redemption price of the plaintiffs' outstanding debentures and the price at which two shares of Boeing's common stock traded on the last day for exercising conversion rights. Id. at 474-75, 100 S. Ct. at 747, 62 L. Ed. 2d at 679. Individual plaintiffs' damages were then determined by the number of debentures each held.

Here, it might be reasonable to presume that all class members, merely by losing power, suffered some damage, if only the inconvenience of having to reset clocks. Even so, departure from the general preference for individualized proof would be warranted only if plaintiffs provided a reliable mathematical formula for calculating aggregate damages. However, in contrast to the plaintiffs in the cases discussed above, plaintiffs here did not offer a formula for calculating, or even estimating, actual damages suffered by class members. Instead they offered analogous reliability value estimates from other utility customers at other times and places [.arrowhorizex] estimates not of actual damages, but of projected damages which might have been suffered in hypothetical circumstances. Judge Perri did not abuse her discretion by rejecting this approach.

b. Fluid recovery theory

Plaintiffs next contend that the court erred by denying the application of the fluid recovery doctrine. Defendants claim that the doctrine did not apply. Part of the class action rule, R. 4:32-2(c), provides, "In any class action, the judgment may, consistent with due process of law, confer benefits upon a fluid class, whose members may be, but need not have been members of the class in suit." The comment to our rule explains this provision:

The significance of the "fluid recovery" principle is its effect on a finding of manageability of the action as a class action where it would be highly impracticable, if not impossible, to allocate and distribute the judgment among persons who were members of the class at the time the action was brought or at the time the judgment was entered.

[Pressler, Current N.J. Court Rules, comment one on R. 4-32, p. 1591 (2004).]

Thus, fluid recovery applies to redistribute "leftover" damages to original class members or others later added to the class, in circumstances where some original class members were proved to be damaged but cannot be located.

In a recent concurring opinion of the Missouri Supreme Court, cited by Judge Perri, one justice explained:

The fluid class recovery doctrine is a rule of equity used in class actions, based on the cy pres doctrine. The fluid recovery doctrine is used for distribution of funds in a class action settlement or award that either cannot be feasibly distributed to individual class members or that remain unclaimed after individual distribution is completed.

[Buchholz Mortuaries, Inc. v. Dir. of Revenue, 113 S.W.3d 192, 196 n. 1 (Mo. 2003) (Wolff, J., concurring).]

A federal appeals court has described the use of fluid recovery in a class action:

Cy pres, or fluid, recovery, is a procedural device that distributes money damages either through a market system (e.g., by reducing charges that were previously excessive), or through project funding (the project being designed to benefit the members of the class). Simer v. Rios, 661 F.2d 655, 675 (7th Cir. 1981). Cy pres recovery "is used where the individuals injured are not likely to come forward and prove their claims or cannot be given notice of the case." Id. at 675. Cy pres recovery is thus ideal for circumstances in which it is difficult or impossible to identify the persons to whom damages should be assigned or distributed.

[Mace v. Van Ru Credit Corp., 109 F.3d 338, 345 (7th Cir. 1997).]

The Mace court concluded that fluid recovery was inappropriate where damages, though small, would not be either difficult to assign or difficult to distribute. 109 F.3d at 345. The Seventh Circuit has explained that the rationale for fluid recovery in class actions is punitive:

In the class action context the reason for appealing to cy pres is to prevent the defendant from walking away from the litigation scot-free because of the infeasibility of distributing the proceeds of the settlement (or the judgment, in the rare case in which a class action goes to trial) to the class members.

[Mirfasihi v. Fleet Mortgage Corp., 356 F.3d 781, 784 (7th Cir. 2004).]

Nothing in the fluid recovery doctrine supports plaintiffs' suggestion that it should substitute for individualized proof of damages merely because known class members may be unlikely to come forward to claim and prove damages. See Mace, 109 F.3d at 345 ("[T]here is no reason, when the injured parties can be identified, to deny them even a small recovery in favor of disbursement through some other means."). Plaintiffs rely onGordon v. Boden, 586 N.E.2d 461, 467 (Ill.Ct.App. 1991), cert. denied, 506 U.S. 907, 113 S. Ct. 303, 121 L. Ed. 2d 226 (1992), for the proposition that fluid recovery is appropriate when "the persons injured are not likely to come forward and prove their claims." In that case, however, the class members were unlikely to come forward because their identities were not known. Ibid. Here, in contrast, their identities are known; the reason they might be unlikely to come forward is that they were not really damaged, were minimally damaged, or could not prove legal damages.

c. Rejection of damages model

Plaintiffs contend that Judge Perri erred by holding that their damages model, in these circumstances, could not serve as a substitute for individualized proof of actual damages. She found that, even if class-wide damage proofs were permitted, the proffered testimony of Billinton and Wacker would be legally insufficient to prove damages, because the surveys of other customer bases could not accurately or reliably represent the actual damages here. We find the judge was correct in rejecting the model proffered by plaintiffs.

As noted, individualized proof of damages is the norm for class actions. Jaroslawicz v. Engelhard Corp., 724 F. Supp. 294 , 300 (D.N.J. 1989). In some circumstances, an econometric model may be used to establish injury on a class-wide basis once the proofs have established that all class members have suffered actual damage. E.g., In re Linerboard, 305 F.3d at 155 (plaintiffs demonstrated through affidavits from expert witnesses that they could prove antitrust impact common to all members of the class). However, in Newton, 259 F.3d at 188, the court emphasized that calculation of the aggregate amount of damages did not absolve the plaintiffs from the duty to prove individual harm.

Plaintiffs contend that there was no basis for excluding their model, which used a method generally accepted in the industry. However, to the extent that the method is accepted in the industry, it is used for entirely different purposes than proving compensatory damages. The testimony of both Billinton and Wacker made clear that the method was designed for, and overwhelmingly used for, cost-benefit analyses to assist system planning decisions, justify new infrastructure, or support tariff changes.

Plaintiffs did not conduct GPU customer surveys of damages following the July 1999 outages, so the court had no occasion to decide if such surveys would have been accepted. However, they cite to the view of industry experts that surveys about hypothetical outages (ex ante) are superior to surveys conducted following an interruption (ex post), which they say the court ignored ("ex ante" assessment means developing a measure before the actual outcome is known, and "ex post" means developing a measure based on what has already happened). Plaintiffs' citations do not fully support their assertion. One of the articles they cite concludes that the two methods of estimating outage costs are both useful and "have a complementary role to play." They also cite deposition testimony in which Billinton expressed that using ex ante studies provided "a valid estimate of the overall costs" incurred and was more reliable than an actual blackout case study of customers actually affected, but only "on aggregate." Billinton emphasized, however, that his and Wacker's report was confined to a "reasonable estimate" of the overall interruption costs, and that they had not "given tremendous amount of thought" to "individual customers and what compensation they would fairly receive."

Plaintiffs point to the certifications of Sullivan and Logan who they say agreed that customer surveys from elsewhere were validly used to calculate damages here. That is not quite our reading of the excerpts plaintiffs cite. Sullivan explained, "Electric utilities use average interruption cost estimates (obtained from population surveys) to evaluate the benefits that can be obtained from investments in service reliability." These estimates can be combined with information on the history of interruptions on circuits within a system. He thought that similar procedures for calculating circuit, customer class and system level interruption costs from historical information "can be used to estimate the costs experienced by different classes of customers on the GPU system." He did not say, however, that survey results could be directly transferred from another customer base in another system.

Nor did Logan say so. He merely said he agreed with the statements of Billinton, Wacker and Sullivan about "the applicability of customer interruption costs obtained by surveys to determining the economic damages incurred by classes of customers resulting from power outages." Also, after listing selected surveys conducted since 1986, Logan said:

Interruption cost surveys such as those above are illustrative of the magnitude of the results that would likely be obtained from surveys for other utilities, and may be used as the basis for reliable estimates of the interruption costs for any other utility if there is a reasonable correspondence between the two utilities in terms of climate, economy, and demographics.

Plaintiffs did not demonstrate correspondences of climate, economy and demographics, nor did they demonstrate other correspondences such as timing and duration of outages, typical usage, or season.

Plaintiffs assert that, contrary to Judge Perri's statement, surveys from one utility's customers are often applied to customers of another utility. The testimony and submissions of plaintiffs' own experts indicates that any such use is limited to estimates of service reliability, not calculation of actual damages, and requires a high level of comparability of circumstances not alleged present here.

Desvousges, defendants' expert, opined that plaintiffs' methodology was improperly applied here fundamentally because the value of reliability, even if accurately measured by the surveys based on hypotheticals, was not a reasonable proxy for actual damages. Nor were the survey results validly transferred. He had used the benefits transfer method in one context only, to measure potential environmental costs associated with utility supply plants. Although he conceded that this method might have been valid here if the surveys had met the necessary criteria of soundness and similarity, he opined that they did not meet those criteria. Moreover, such a benefits transfer method would have provided only a "starting point" to estimate damages, because of the inevitable error associated with such transfer.

The larger point Judge Perri was making, moreover, was that ex ante customer surveys, whether from the same utility or another one, have never been found valid for, or ever used for, the purpose of awarding litigation damages. The judge said:

Indeed, this method of calculating damages has never been applied to providing compensation to customers following an actual power outage nor has it been used to calculate the supposed losses sustained by customers after an actual electricity outage, with the exception of two isolated cases in Norway and Sweden. Rather, this method of calculating damages has only been addressed within the context of infrastructure planning. No peer review papers have addressed how, when or whether this data can be used to calculate actual losses after an actual blackout. Finally, plaintiffs' experts acknowledge this is the first and only time where survey results have been taken from one utility company's customers and then applied to a different utility company.

Indeed, Billinton conceded in deposition that reliability worth analysis had been applied only to entire systems, never to individual customers. Wacker also said that he knew of no one who had used such cost interruption data to estimate actual losses incurred, nor had he done so.

Plaintiffs contend that whether results from customer damage surveys transferred from other customer bases should be used was a question for a jury to decide, and that the accuracy of such transfers went to the weight, not the admissibility, of the evidence. We disagree. The decision to admit or exclude expert testimony lies within the sound discretion of the court. State v. Moore, 122 N.J. 420, 459 (1991).

The standard for admissibility of expert testimony is set forth in N.J.R.E. 702:

If scientific, technical, or other specialized knowledge will assist the trier of fact to understand the evidence or to determine a fact in issue, a witness qualified as an expert by knowledge, skill, experience, training, or education may testify thereto in the form of an opinion or otherwise.

New Jersey imposes a three-part test for the admission of expert testimony:

(1) the intended testimony must concern a subject matter that is beyond the ken of the average juror; (2) the field testified to must be at a state of the art such that an expert's testimony could be sufficiently reliable; and (3) the witness must have sufficient expertise to offer the intended testimony.

[State v. Kelly, 97 N.J. 178, 208 (1984).]

Here, Judge Perri found that the proffered testimony met the first and third requirements, but not the second. She found that the transferred customer survey results were not sufficiently reliable as a measure of damages because they were not designed to measure, and had never been used to measure, actual damages. In addition to the inherent lack of correspondence with the circumstances of the outages here, the surveys themselves focused on the value the customers placed on the reliability of their electrical service, not actual damages.

Plaintiffs contend that they showed general acceptance of their experts' methods, and the reliability of the customer survey method. Even assuming they did, however, they did not show that the method was reliable when the results were applied to make assumptions about other customers in other locations. Especially where there was a lack of correspondence of seasons, climate and demographics, there was no presumed correspondence of the projected value of lost use with the actual damages suffered here. Nor did plaintiffs show that the method was reliable to estimate actual damages suffered, as opposed to valuations of prospective losses. Thus, even accepting plaintiffs' assumption that they showed the reliability of the survey results, they failed to show that the use of those results here constituted a valid measure of damages.

Wacker admitted that a study of GPU customers after the event, using either questions about a hypothetical event or questions particular to this outage, would have produced the best survey data. However, he and Billinton concluded that relating the interruption costs to the specific occurrences would be difficult, though not impossible. A sample survey alone would cost several hundred thousand dollars.

We agree with defendants that the issue is not whether it was "unrealistic" to expect plaintiffs to present individual proof of damages, as plaintiffs say, but whether the model they proffered was a permissible representation of damages. One reason the model did not satisfy that standard was because the expert testimony was not reliable in context. Plaintiffs point to no other case where a court has accepted a similar proof of damages by analogy.

"[T]he usefulness of expert testimony depends in part on the context in which it is offered." State v. Cavallo, 88 N.J. 508, 526 n. 8 (1982). Because expert testimony must assist the trier of fact, its admissibility depends in part on the connection between the evidence to be presented and the disputed factual issues in the case. In re TMI Litig., 193 F.3d 613, 665 (3d Cir. 1999).

Although the judge explained her rejection of the transferred survey evidence based on its unreliability in this context, that rejection could also be viewed as a finding that the evidence was not sufficiently relevant to warrant admission because it was neither material nor probative. Relevant evidence is ordinarily admissible. N.J.R.E. 402. "`Relevant evidence' means evidence having a tendency in reason to prove or disprove any fact of consequence to the determination of the action."N.J.R.E. 401. The determination of relevancy focuses on "the logical connection between the proffered evidence and a fact in issue." State v. Hutchins, 241 N.J. Super. 353, 358 (App.Div. 1990). One court explained:

Relevancy is composed of two parts: materiality and probative value. McCormick on Evidence, § 185 at 541 (3rd ed. 1984). "Materiality looks to the relation between the propositions for which the evidence is offered and the issues in the case." Ibid. "Probative value" is "the tendency of evidence to establish the proposition that it is offered to prove." Ibid. Direct evidence "offered to help establish a provable fact can never be irrelevant" but circumstantial evidence may be "so unrevealing as to be irrelevant." Id. at 543-544. "[T]o say that evidence is irrelevant in the sense that it lacks probative value is to say that knowing the circumstantial evidence does not justify any reasonable inference as to the fact in question." Id. at 544. A determination of relevancy lies in "the judge's own experience, his general knowledge, and his understanding of human conduct and motivation." Ibid.

[State v. Allison, 208 N.J. Super. 9, 17 (App. Div.), certif. denied, 102 N.J. 370 (1985).]

The trial court is granted broad discretion in determining the relevance of evidence. Green v. New Jersey Mfrs. Ins. Co., 160 N.J. 480, 492 (1999).

Here, the judge essentially found that the survey evidence lacked materiality, because it did not pertain to the actual damage sustained by class members, and lacked probative value, because it was not similar enough to the actual damage to justify the desired inference.

d. Other defects in damages model

Defendants make two additional arguments to support the rejection of plaintiffs' damages model. In our view Judge Perri correctly found it unnecessary to reach these arguments, and we discuss them only briefly.

The first argument is that, because the model consists of extrapolation of information from surveys, it is inadmissible hearsay. N.J.R.E. 801(c). Plaintiffs do not deny that the surveys themselves constitute inadmissible hearsay. They respond that they do not seek to have them admitted into evidence, although their experts intend to rely on them.N.J.R.E. 703 provides in part, "If of a type reasonably relied upon by experts in the particular field in forming opinions or inferences upon the subject, the facts or data [on which an expert relies] need not be admissible in evidence." Nevertheless, "it is not a rule of wholesale admissibility of otherwise inadmissible evidence." State v. Raso, 321 N.J. Super. 5, 16 (App.Div.), certif. denied, 161 N.J. 332 (1999).

If an expert relies on the truth of the matter asserted in formulating an opinion, the jury should be instructed that the probative value of the opinion is dependent upon that. State v. Vandeweaghe, 351 N.J. Super. 467, 480-81 (App.Div. 2002), aff'd, 177 N.J. 229 (2003). We agree with defendants that here, the experts are relying so thoroughly on the surveys that they would be merely summarizing and interpreting hearsay.

Defendants' second argument is that the model denies them their due process right to have the trier of fact determine all issues after they had the opportunity for cross-examination on the claims of all class members. Plaintiffs respond that they produced much of the underlying Canadian surveys and hundreds of pages of materials regarding the PGE surveys, which provide ample basis to cross-examine. However, opportunity to cross-examine concerning the surveys is not meaningful if, as defendants argue and the court found, the survey results do not represent actual damages here.

In two asbestos cases, Cimino v. Raymark Industries, Inc., 151 F.3d 297, 319-21 (5th Cir. 1998), and In re Fibreboard Corp., 893 F.2d 706, 711-12 (5th Cir. 1990), the courts rejected a procedure in which damages payable to over 3000 plaintiffs would be determined by extrapolating from a representative sample of class members. It would include a full trial of liability and damages for eleven class representatives, along with evidence from thirty illustrative plaintiffs. Id. at 708-09. Of this proposed procedure, the Fifth Circuit said in Fibreboard, "It is called a trial, but it is not." Id. at 712. The court found the procedure violated Texas law and went beyond the judicial authority, and that the defendants' concerns "f[ou]nd expression in defendants' right to due process." Id. at 711.

Plaintiffs counter that defendants' only due process interest is in the total amount they must pay, not in its allocation among individual claimants. They rely on Hilao v. Estate of Marcos, 103 F.3d 767, 786 (9th Cir. 1966), and In re Simon II Litigation, 211 F.R.D. 86 , 104, 153 (E.D.N.Y. 2002). These cases are distinguishable primarily because here, plaintiffs have not presented a reliable formula for determining aggregate damages.

In Hilao, 103 F.3d at 784-86, the court explained that use of a representative sample to determine what percentage of the total claims were valid (claims by or on behalf of victims of human rights abuses by former Philippines president Ferdinand Marcos) did not violate due process. Although the risk of error in a statistical method was somewhat greater than in an adversarial adjudication of each claim, the defendants were protected because the proof of claim form required to be submitted by each class member included a certification that the information was true and correct. Id. at 786. Here, in contrast, no such safeguards are present. More important, here, the statistical method itself lacks reliability. Note that the court in Cimino, 151 F.3d at 319, expressly disagreed with Hilao, and agreed with its dissenting opinion, on this issue.

In Simon II, sufferers from smoking-related illnesses sued cigarette manufacturers for consumer fraud, and the court approved the use of statistical evidence to establish causation and damages.211 F.R.D. at 146-47. The court concluded that the use of aggregate proof did not violate the defendants' due process and jury trial rights; without it the case would be impossible to try. Ibid. The court found that in that case, statistical analysis might "provide a more accurate and comprehensible form of evidence than would the testimony of millions of individual smokers." Id. at 148. Here, in contrast, the court found that the proffered aggregate proofs were unreliable.

The Simon II court acknowledged that the Fifth Circuit had rejected the use of aggregate statistical techniques in class actions where damages needed to be apportioned among individuals, despite approving them to calculate lump sum damages in mass tort cases. Id. at 156. It also acknowledged that the defendants had an interest in "not paying for damages in excess of what alleged misconduct may have caused," and that interest would be furthered by confronting each individual. Id. at 153. However, the impossibility of that approach tempered the weight of the tobacco companies' interest. Ibid. The court concluded that it was unreasonable not to rely on statistical evidence in a mass tort class action, especially where, as was alleged, the defendants themselves had made a claim-by-claim showing virtually impossible "by distorting the entire body of public knowledge on the lethal and addictive effects of smoking." Ibid. Concerning punitive damages, the court noted that their "exact division among class members" was not critical for the defendants as long as the total sum was fair. Id. at 104.

Here, in contrast, individual proof of damages, through customer claim forms, is not impossible or even impractical. Unlike the circumstances ofSimon II, here defendants are not responsible for impeding the individualized proof of damages. Moreover, unlike the tobacco cases, where each class member would have to produce the same kind of statistical proof on the effects of smoking, here the customers' damages are dependent instead on their individual circumstances. We find the court did not err by rejecting plaintiffs' proposed proofs of classwide damages.

With respect to proof of damages on the remand, as to the "Red Bank" class, the proof or disproof of damages in individual cases should be feasible through use of customer claim forms and surveys, judicious use of interrogatories and demands for admission, reasonable investigative efforts, and perhaps statistical interpretation of sampling data from the relevant universe, established based on competent data. We do not in any sense intend to limit the demonstrated thoughtful creativity of the trial judge, and the manifest energy of the parties in this regard.

We affirm the orders decertifying the class and denying the motions to permit expert testimony on proposed class-wide damages but we remand for certification on the more limited Red Bank class, i.e., that class of customers whose outages directly resulted from the alleged negligence in delaying the replacement of the transformers at the Red Bank substation.