United States of America et al v. NCR Corporation et al

Filing 1033

ORDER signed by Chief Judge William C Griesbach on 10-19-15 granting 999 Motion for Order; granting in part and denying in part 1003 Motion for Reconsideration ; granting 1005 Motion for Reconsideration ; granting 1007 Motion to Intervene; granting 1010 Motion for Reconsideration ; granting 1013 Motion for Reconsideration ; granting 1025 Motion for Leave to File; granting 1027 Motion for Leave to File. (cc: all counsel) (Griesbach, William)

UNITED STATES DISTRICT COURT
EASTERN DISTRICT OF WISCONSIN
UNITED STATES OF AMERICA and
STATE OF WISCONSIN,
Plaintiffs,
v.

Case No. 10-C-910

NCR CORP., et al.,
Defendants.

ORDER GRANTING RECONSIDERATION

Following this Court’s May 15, 2015 Decision and Order finding that NCR had established
its divisibility defense, most of the other parties filed motions seeking reconsideration. These
parties argue that NCR failed to establish its divisibility defense because the evidence and testimony
that are the subject of the limited remand were unreliable. For the reasons given below, I agree, and
will therefore grant the motions.
I. Motions for Reconsideration
1. The Harm is not Theoretically Capable of Apportionment
As the Seventh Circuit explained, the harm is “theoretically capable of apportionment if
NCR could show the extent to which it contributed to PCB concentrations in OU4.” United States
v. P.H. Glatfelter Co., 768 F.3d 662, 678 (7th Cir. 2014). As noted in my previous decision (ECF
No. 1000), the government appeared to have all-but conceded this point, but in its subsequent
motion practice it has launched a vigorous assault on Dr. Wolfe’s estimates, which the Seventh
Circuit has asked this Court to consider on remand.

It is true, as NCR notes, that some of the issues now raised could have been argued at some
earlier time. In some circumstances, many of the arguments would not properly be the subject of
a motion for reconsideration. Here, however, the circumstances are highly unusual. Following the
remand, the government filed a motion, accompanied by a short brief, asking the court to conclude,
based on the trial record, that NCR had failed to meet its burden. (ECF No. 935.) NCR responded,
and the government filed a reply brief. NCR filed a surreply brief. During the limited briefing on
that motion, the numerous other interested parties were not heard in any meaningful capacity.
Neither were the other parties (nor the government) able to meaningfully respond to NCR’s
approach during trial, because Butler’s use of Wolfe’s evidence came only in NCR’s surrebuttal.
Given the import of the matters under consideration, the motions for reconsideration allowed these
parties to be heard and to provide further context, a context that makes it clear that NCR’s lateinning use of Dr. Wolfe’s evidence was not sufficient to meet its burden.
The most salient problem with Wolfe’s estimates is that they contradict facts already found
by the court. For example, several of the parties note that Wolfe’s estimates overstate—probably
dramatically—the PCBs attributed to U.S. Paper, one of the four key sources of PCB contamination,
which is located in the upper half of OU4, known as OU4A. As a point of reference, following the
trial this Court found that NCR’s expert, James Braithwaite, had relied on an excessive estimate of
some 9,189 kg attributable to U.S. Paper. In rejecting that conclusion, I noted that U.S. Paper had
not used NCR broke during much of the PCB era, and when it did use NCR broke, it used only
small amounts. (ECF No. 794 at 38-39.) Specifically, this court concluded, as a finding of fact, that
U.S. Paper “did not recycle NCR paper, which was white, as a significant part of its business. In
about 1966, however, it began manufacturing white paper rolls using white NCR broke, but even
2

then it used only a small portion of NCR broke in the process, as white rolls themselves constituted
only a limited portion of its recycled product business.” (Id. at 15.) Another expert, David Merrill,
had more credibly estimated that U.S. Paper had discharged no more than 1,166 kg of PCB, which
would be somewhere in the neighborhood of one-half of one-percent of the total PCBs released to
the river. (Tr. 2320-31.) In sum, NCR’s expert had attributed 9,189 kg to U.S. Paper, while a more
credible expert had found no more than 1,166 kg.
In contrast, Georgia-Pacific’s expert, Dr. Wolfe, estimated that U.S. Paper had released
some 21,500 kg of PCBs, a figure that could easily be too high by a factor of nearly twenty. That
explains why he believed U.S. Paper (the only significant OU4A-based source of PCBs) would have
been responsible for some 21% of the PCBs in OU4B, and an even higher percentage in OU4A
itself, even though U.S. Paper likely contributed only a tiny fraction of the total PCBs to the
river—at most, in the low single digit percentages. Obviously, Wolfe’s estimate is sharply at odds
with that reality. (Notably, NCR does not deny the essence of that fact.) Because there are only a
few sources of contamination into OU4, a large error with respect to a single source will skew the
numbers dramatically in favor of the other sources, including NCR.
This calls into doubt both the general framework used by Dr. Wolfe as well as his specific
conclusions. Most problematic is the fact that his estimates for NCR were not unfavorable or
conservative, as I had previously concluded, but instead were based on the flawed premise that U.S.
Paper was a substantial polluter of PCBs into the river. As discussed below, Wolfe did not explain
exactly what percentage he attributed to U.S. Paper for OU4A (his expert report was limited to
conclusions about OU4B), but it would have been much higher than the 21% he ascribed to U.S.
Paper for OU4B. That means Wolfe’s conclusion attributing some 43% of the PCBs in OU4A to
3

NCR was actually low, not high. If much lower assumptions are made with respect to U.S. Paper,
then it follows that NCR could easily have been responsible for 50% or more of the harm in OU4A
and some 35% or more in OU4B. We do not know. The point is, this substantial divergence from
facts I have found, and which are not presently contested, calls Wolfe’s conclusions into serious
doubt, so much so that NCR has not met its burden to “show the extent to which it contributed to
PCB concentrations in OU4.” Id. at 678. In short, Butler’s use of Wolfe’s flawed estimates cannot
be expected to produce a reliable answer to the question of how much NCR contributed to PCB
concentrations in OU4.
2. Remediation Costs Necessitated by Each Party
The parties spend the bulk of their efforts addressing the second divisibility question, which
is whether there is a reasonable basis for apportionment. Once again, the further context provided
in subsequent briefing has shed light on some glaring flaws in NCR’s approach.
A. Missing Estimates
In its motion for reconsideration, the moving parties argue that the trial record does not
provide complete estimates for all of the dischargers, and so Butler’s use of those estimates must
be based on mere guesswork rather than supportable science. The moving parties focus on OU4A,
the artificial designation for the upstream half of OU4. Wolfe’s expert report provided PCB
concentration estimates for OU4B (the location of Georgia-Pacific, his client), but not for OU4A.
During the trial, however, it came out that Wolfe had produced an estimate of NCR’s contribution
to OU4A (43%), and that is how Butler learned that figure, which he subsequently used in his own
model and testified about on surrebuttal. In sum, Wolfe provided his estimates for four of the key
sources of pollution: the OU1, OU2 and OU4 contributions to pollution in OU4B, as well as the
4

43% estimate for OU2 (NCR) contribution to pollution in OU4A.
But, the moving parties note, there was no evidence as to Wolfe’s estimates about two of
the other key sources of pollution into OU4A, namely those in OU1 (like Glatfelter) and OU4A
itself (U.S. Paper). Accordingly, because Wolfe’s estimates were incomplete, Butler could not have
arrived at a meaningful stand-alone estimate because he did not have estimates for all of the key
PRPs. He must simply have guessed at them.
NCR protests that there is no mystery as to where Butler’s data came from because it is
readily discernable from Wolfe’s expert report. The parties debate whether the expert report was
itself admitted into evidence, but that is irrelevant. An expert is not bound to rely only on
“evidence” admitted at trial. Whether Wolfe’s report was admitted as evidence or not, NCR argues,
Butler read it and relied on it. (Tr. 1514, ECF No. 726.) Even if true, the report still limits itself
to estimates for OU4B rather than OU4A. (ECF No. 628-9 at 63-64.) That is, there is no table or
other graphic demonstrating Wolfe’s conclusions about OU4A. Thus, contrary to this court’s earlier
belief, it was not simply a matter of Butler using another expert’s mass estimates and running them
through his apportionment model. Instead, two of the key estimates are missing from the data set
that Butler relied upon in reaching his conclusions.
NCR argues that the estimates are “readily obtained” from Wolfe’s expert report, (ECF No.
1025-1 at 3.), but its choice of the passive voice verb is telling. It is conceivable that Wolfe’s
estimates for the other operable units from data contained within the report are obtainable, or that
they are “obtained” now, but Butler did not testify that he ever actually obtained them, nor did he
explain how he would have done it. The fact that a report might allow an expert to divine another
expert’s estimates does not mean that the expert actually did so.
5

In fact, the trial testimony suggests that Butler did not have complete data from Wolfe,
contrary to this court’s earlier assumptions. Recall that Butler did not have any of Wolfe’s OU4A
estimates until he was cross-examined by Georgia-Pacific’s counsel, who informed him that Wolfe
indeed had an estimate for OU2 (NCR) contributions to OU4A. That was the only OU4A estimate
Butler had. As NCR’s counsel explained during the trial, Butler “has since [i.e., since the crossexamination] been able to use his model and use that input, which he had not had an opportunity
to do until our rebuttal evidence.” (ECF No. 731 at 141, Tr. 2782.) And Butler himself testified
that Wolfe’ report “did not provide an estimate for OU4A.” (Id. at 139, Tr. 2780.) If Butler had
actually found and used all of Wolfe’s key estimates in the expert report, he would not have needed
to wait until trial to use Wolfe’s OU2 estimate for OU4A—he would already have known it. As
NCR’s counsel explained at trial in response to an objection, “all he's done [Butler] is redo exactly
the same model that counsel have had for weeks or months, simply changing one variable, which
is the 43 percent estimate, in place of the 37 percent Simon estimate that he used the last time.” (Id.
at 2784.) Thus, it seems abundantly clear that Wolfe used the 43 percent estimate for NCR, but did
not use Wolfe’s other estimates for OU1 and OU4. He only changed “one variable”—Wolfe’s
estimate of 43 percent for OU2 contributions into OU4A.
In short, if Butler had actually obtained the estimates for OU4A in the report, he would have
said so. Instead, it is clear that NCR is now attempting to reconstruct what could have been done
rather than explaining what Butler actually did. In short, it is largely a mystery as to how Butler was
able to arrive at a 28% figure for NCR’s contribution to the harm in OU4 because Butler was
missing two of Wolfe’s key estimates for OU4A. Instead, it appears that Butler’s model
incorporated and relied upon incomplete estimates, and, as I found above, the estimates from Wolfe
6

that Butler did have were themselves deeply flawed. Accordingly, I now conclude that Butler’s use
of Wolfe’s estimates did not provide a reasonable basis for apportioning the costs.
B. Changed Paradigm
A second reason for rejecting Butler’s apportionment approach is that he was operating
under a paradigm (what we have been calling the “binary” approach) that is no longer operative.
Butler’s approach to apportioning costs produced results that changed NCR’s share very little, even
if the PCB mass assumptions varied widely. For example, when Butler used Wolfe’s 43% figure
instead of the 37% figure he had adopted from the Simon team, the share he now attributed to NCR
rose from 23.5% to 23.7%, a measly two-tenths of one-percent, even though Wolfe’s 43% figure
was some 16 percent higher (6 / 37ths) than 37 percent. Under Butler’s method, one could ascribe
to NCR very large mass estimates without moving the needle very much. Butler explained this
apparent anomaly by citing the now-abandoned binary model. The reason there was such a small
change in NCR’s share, he explained, “has to do with the relationship between PCB mass and cost,
which is nonlinear. And that flows from the Remedial Action Decision Rules and the very low 1
part per million PCB threshold here. So that when we're in a situation like we are in Scenario 4 with
37 percent, we're already a fair amount above the 1 part per million threshold. So by adding
additional mass and increasing the concentration, what you see is a very small change in the
footprint of the remedy, very small change in the remedial quantities, and a very small change in
the cost.” (Id. at 2785-86.) Almost every assumption Butler relied on was overruled by the Seventh
Circuit, which held:
[T]he harm resulting from PCB contamination in the Lower Fox River cannot be
characterized as binary. PCB concentrations below the 1.0 ppm remedial action level
and even the 0.25 ppm target SWAC still pose a threat to human health and the
7

environment. Even in areas where no remediation is required, higher PCB
concentrations contribute more to the risk of harm and require a longer period of
natural recovery to achieve an acceptable risk. Moreover, not all concentrations
above EPA's remedial thresholds are equally harmful; the risk of harm increases
with concentration even at high levels.
The continuous (as opposed to binary) nature of PCB contamination in the Lower
Fox River leads us to reexamine EPA's remediation rules to determine whether
remediation costs are still a useful approximation of the contamination caused by
each party. As with the contamination, the district court thought that remediation
costs resembled an on/off switch: sediment with PCB concentrations below 1.0 ppm
would impose no remediation costs, while sediment with PCB concentrations above
1.0 ppm would always impose about the same remediation costs. We think the
district court got this wrong as well. In fact, remediation costs increase with the
degree of contamination above 1.0 ppm. As a result, remediation costs are still a
useful approximation of the degree of contamination caused by each party.
United States v. P.H. Glatfelter Co., 768 F.3d 662, 677 (7th Cir. 2014).
In short, key assumptions in Butler’s apportionment model caused it to be insensitive to
additional PCB mass estimates, in contradiction of the Seventh Circuit’s view that both the harm
and the cleanup costs are relatively linear, i.e., non-binary. These assumptions were fundamental
to Butler’s approach, and as such they infect his entire model. Because the fundamental premise
of his model has been overturned on appeal, the entire model cannot be used as a sound basis for
apportioning costs.
3. Causation and Equity
Glatfelter and others argue that since NCR caused the entirety of the harm by producing
carbonless copy paper in the first place, the harm must not be divisible. But this effort ignores that
equity has no role in divisibility, and accepting it would create a kind of product liability framework
for CERCLA rather than one based on actual pollution, which is a strict liability tort. In every
pollution case, the pollution would not have occurred but for the creator of the toxin—the chemical,

8

oil, or whatever. Here, that company is Monsanto, which is not even a party. CERCLA, however,
operates on a pollution model, not a products liability model, and so to the extent any of the motions
rely on that argument, they are denied.
II. Motion to Withdraw Sixth Claim for Relief
The Plaintiffs also filed a motion seeking leave to withdraw the sixth claim for relief in their
first amended complaint. They note that the claim, which seeks natural resource damages, is being
withdrawn in connection with the receipt of some $45.9 million in settlements. Glatfelter has
responded to the motion by asking that any dismissal of the claim be with prejudice rather than
without.
A court may, in its discretion, allow a litigant to withdraw a claim so long as the motivation
is not based on procedural gamesmanship. Solaia Tech. LLC v. Arvinmeritor, Inc., No. 02-C-4704,
2004 WL 2203437, at *6 (N.D. Ill. Sept. 29, 2004). Here, the government has offered ample
explanation for no longer wishing to pursue the claim in the context of this litigation. Functionally,
the “withdrawal” of a claim is not a dismissal but an allowance of an amended complaint. Glatfelter
has not presented any sound basis to conclude that the government’s withdrawal of the complaint
should be treated as a dismissal with prejudice. Accordingly, the claim will be considered
withdrawn.
III. Conclusion
For these reasons, the motions for reconsideration are GRANTED. I conclude that NCR
has failed to meet its burden to demonstrate both that the harm is theoretically capable of divisibility
and that there is a reasonable basis for apportionment. The United States is directed to submit a
proposed judgment within seven days.
9

The motion to intervene [1007] is GRANTED. The motions to file a sur-reply [1025, 1027]
are GRANTED.
The motion for leave to withdraw [999] is GRANTED. The Plaintiffs’ Sixth Claim for
Relief (seeking natural resource damages) in their First Amended Complaint shall hereby be deemed
withdrawn at Plaintiffs’ request.

SO ORDERED this 19th day of October, 2015.
/s William C. Griesbach
William C. Griesbach, Chief Judge
United States District Court

10