Safeco Insurance Co. of America v. American International Group, Inc.

In the
United States Court of Appeals
For the Seventh Circuit
Nos. 12-1157, 12-1158, 12-1186,
12-1730, 12-1753 & 12-1764
S AFECO INSURANCE C OMPANY OF A MERICA, et al.,
Plaintiffs-Appellants,
and
L IBERTY M UTUAL INSURANCE C O ., et al.,
Counterclaimants-Appellants,
and
ACE INA H OLDINGS, INC., et al.,
Intervening Plaintiffs-Appellees/
Defendants-Appellees,
v.
A MERICAN INTERNATIONAL G ROUP, INC., et al.,
Defendants-Appellees/Plaintiffs-Appellees/
Counterdefendants-Appellees.
Appeals from the United States District Court
for the Northern District of Illinois, Eastern Division.
Nos. 07 C 2898 & 09 C 2026—Robert W. Gettleman, Judge.
2 Nos. 12-1157, et al.
A RGUED N OVEMBER 29, 2012—D ECIDED M ARCH 25, 2013
Before E ASTERBROOK, Chief Judge, and P OSNER and
M ANION, Circuit Judges.
E ASTERBROOK, Chief Judge. About 45 days after these
appeals had been argued, the appellants asked us to
dismiss them, see Fed. R. App. P. 42(b), informing the
court that the dispute had been settled. All but one of
the appellees (ACE INA Holdings) joined a stipulation
of dismissal; ACE did not join it, but neither does it
oppose dismissal. Because the litigation is a class action,
however, we were concerned that the settlement might
have adverse effects on other members of the class. So
we asked for additional memoranda. These have been
filed, and in the two months that have elapsed since the
notice no member of the class has expressed opposi-
tion. Having concluded that the settlement does not
jeopardize the interests of the unrepresented class mem-
bers, we dismiss the appeals.
The first sentence of Rule 42(b) provides that, if all
parties agree to an appeal’s dismissal, then the clerk of
court may close the proceeding without judicial action.
ACE did not join the stipulation, so the second sentence
of Rule 42(b) applies: “An appeal may be dismissed on
the appellant’s motion on terms agreed to by the parties
or fixed by the court.” This sentence uses “may” rather
than “must” so that the judges can protect the rights
of anyone who did not consent to the dismissal. Although
Nos. 12-1157, et al. 3
the members of the class are not technically parties,
they have legally enforceable interests. See Devlin v.
Scardelletti, 536 U.S. 1 (2002).
Companies underwriting workers’ compensation insur-
ance participate in a reinsurance pool administered
by the National Workers Compensation Reinsurance
Association (the Association). Insurers share in the pool’s
profit or loss according to the volume of business they
underwrite. When the pool is profitable, it is beneficial
to have a larger book of business; when the pool loses
money, a smaller book means that the underwriter
needs to contribute less toward the losses. The class in
this suit contends that American International Group
(AIG) underreported the size of its business in losing
years, causing the pool’s other members to bear a
disproportionate share of the losses. The class asked
for about $3.1 billion.
Some of the insurers had other business dealings.
Liberty Mutual and its affiliates, including Safeco, have
independent claims against AIG. For its part, AIG ad-
vanced claims against Liberty Mutual (as we call the
entire group). When Liberty Mutual caused Safeco to
commence this class action as the representative plain-
tiff, these other claims complicated the litigation. Once it
became evident that Liberty Mutual had unacceptable
conflicts, ACE INA Holdings intervened, with several
other insurers, to take over as the class’s representatives.
Still, Liberty Mutual sought to use the class suit as a
club to induce AIG to pay more on its separate claims
against AIG, while AIG sought to minimize the sum
4 Nos. 12-1157, et al.
of what it paid the class plus what it owed Liberty
Mutual separately.
ACE (and the other new representatives, which we
ignore from here on) eventually settled the class claims
against AIG for $450 million. The settlement includes
releases of all claims that pool members held against AIG
in all lines of business (not just reinsurance of workers’
compensation policies), plus releases of AIG’s claims
against the class’s members. Liberty Mutual protested;
it contended that its 22% share of the settlement
(some $99 million) is too small, given the value of its
independent claims against AIG. The settlement pro-
vides that any class member can opt out, and ACE antici-
pated that Liberty Mutual would do so. The settlement
agreement provides that, if Liberty Mutual were to opt
out, AIG’s payment would be reduced to $351 million.
Liberty Mutual elected to stay in the class. So did all
but one other insurer. The district judge approved the
settlement after a hearing under Fed. R. Civ. P. 23(e). 2012
U.S. Dist. L EXIS 25265 (N.D. Ill. Feb. 28, 2012). Liberty
Mutual then appealed, arguing in this court that its
share of the settlement does not compensate it
adequately for the value of its stand-alone claims
against AIG. It also contended that the conflicts of
interest within the reinsurance pool meant that the case
never should have been certified as a class. (This argu-
ment appears in Safeco’s brief rather than Liberty Mu-
tual’s, but as they are under joint control the main effect
of filing separate briefs is to get extra words. None of
the other parties contends that Safeco should be viewed
Nos. 12-1157, et al. 5
as independent of its parent; after all, this is why Safeco
was not a satisfactory class representative.) Appellants
made some other arguments, which need not be de-
scribed. None of the insurers outside the Liberty Mutual
group complained about the class certification or the
settlement, and the Association, on behalf of the entire
pool, supported the district court’s decision.
After argument, Liberty Mutual settled with AIG.
The terms of the settlement do not matter to the other
members of the class, who still split $351 million among
them. ACE and the other representatives are content.
Neither the Association (which manages the pool) nor
any member of the class has protested. It is accordingly
hard to see how a live controversy remains, and courts
should not issue opinions resolving litigation that
the parties no longer want to pursue. Since no one now
wants us to adjudicate this dispute—or even suggests
that there is a “dispute” left to adjudicate—dismissing
the appeals is in order. Cf. U.S. Bancorp Mortgage Co. v.
Bonner Mall Partnership, 513 U.S. 18 (1994).
We have considered, in the spirit of Rule 23(e),
whether this settlement has any potential to injure non-
participants. Yet all of the pool’s members outside the
Liberty Mutual group still get exactly what they
accepted before—and the district court found that reso-
lution fair. Liberty Mutual’s appeal principally con-
cerns the way the district court’s order affects its own
claims against AIG. That’s something Liberty Mutual
had every right to resolve independently by opting out.
A settlement between Liberty Mutual and AIG while
6 Nos. 12-1157, et al.
the appeal was pending works as a belated opt-out,
which has no greater potential to injure the pool’s other
members than an opt-out before the district court acted
would have done. If, under the settlement, opt-out by
Liberty Mutual meant undoing the pact and continuing
the litigation, then a de facto opt-out on appeal might
justify a remand. But the possibility of Liberty Mutual
opting out and reaching a side deal with AIG was pro-
vided for in the settlement itself. That this possibility
now has been realized does not call into question the
settlement’s fairness to the pool’s other members.
Could Liberty Mutual’s appeal itself have injured
other members of the class—perhaps by leading them to
think that they needn’t file their own appeals? That is
very unlikely, for three reasons.
First, what issues would other class members have
raised on appeal? None had complained about the settle-
ment, so there was no adverse decision to appeal from.
Second, why would Liberty Mutual’s appeal have dis-
suaded another insurer from appealing? Any other firm
could see that Liberty Mutual was appealing to defend
its separate interests; none would have relied on Liberty
Mutual. The established conflict between Liberty Mutual
and the rest of the class is why ACE intervened to take
over as the representative plaintiff. Other members of the
class would have seen Liberty Mutual as a threat to their
interests, not as a champion they could rely on for protec-
tion. Recall that Liberty Mutual asked us to abrogate the
class certification, a step that would have eliminated the
other insurers’ recoveries.
Nos. 12-1157, et al. 7
Third, Liberty Mutual waited until the end of the win-
dow for appeal. The judgment was entered on Feb-
ruary 28, 2012, and the appeals were filed on March 27.
No other insurer could have been safe in waiting to
see whether Liberty Mutual would appeal; a pool
member that wanted appellate review would have
acted on its own before March 27. Liberty Mutual would
not have violated any other insurer’s rights by settling
with AIG on March 26 and never filing an appeal; filing
an appeal at the end of the available time and settling
later has no greater potential to injure other members
of the class.
Although we appreciate that conflicts of interest
between representative plaintiffs and class members
can lead the representatives to sell out for too little, no
one has accused ACE of yielding to that temptation.
All members of the class are large and sophisticated
businesses, many with millions on the line and
legal staffs to protect their interests. Even the smaller
insurers receive more than $100,000 from the settle-
ment, and if the representatives had been able to
negotiate for the $3 billion the class initially sought,
the average return per insurer would have exceeded
$2 million (and about $750,000 apiece for the smaller
insurers). The pool is a multi-billion-dollar business;
its manager, which looks out for the aggregate of all
members’ interests, supports both the original settle-
ment and the dismissal of Liberty Mutual’s appeal. The
Manual for Complex Litigation §21.61 (4th ed. 2004),
provides a list of events that may tip off the judiciary to
8 Nos. 12-1157, et al.
a problem; none of the things to watch for has occurred
in this suit.
Because there is no prospect of injury to any other
class member, we need not discuss at length this state-
ment in the committee note to the 2003 amendment to
Rule 23(e): “Once an objector appeals, control of the
proceeding lies in the court of appeals. The court of
appeals may undertake review and approval of a set-
tlement with the objector, perhaps as part of appeal
settlement procedures, or may remand to the district
court to take advantage of the district court’s famil-
iarity with the action and settlement.” The com-
mittee note does not discuss any particular language
in Rule 23, which like the other civil rules deals with
proceedings in district courts rather than courts of ap-
peals. All the committee’s statement does is recognize
that the court of appeals will decide what to do.
For the reasons we have given, we do not think any
further proceedings necessary.
If despite appearances this settlement makes other
class members worse off or disappoints their reasonable
expectations, a class member could file a motion in the
district court under Fed. R. Civ. P. 60(b)(3) (misconduct
by an opposing party) or 60(b)(6) (“any other reason
that justifies relief”). If such a motion were to be filed,
a concrete controversy would call for judicial resolu-
tion. At the moment, however, none of the parties
wants to fight, and none of the class members has ex-
pressed dissatisfaction. Any further proceedings would
be gratuitous. The appeals are dismissed.
Nos. 12-1157, et al. 9
P OSNER, Circuit Judge, dissenting. Dismissal of the
appeal in this class action suit is premature. It is based
on speculation rather than on evidence, is insensitive
to the risks of class action sell-out, and makes critical
errors.
We don’t know the terms of the settlement on which
dismissal is predicated, so we don’t know whether the
settlement sells out the interests of the class. But it
may. In discounting that possibility the majority
opinion makes two critical errors. The first is to say
that “any other firm [that is, any other class member]
could see that Liberty Mutual was appealing to defend
its separate interests; none would have relied on
Liberty Mutual.” Yet two subsidiaries of Liberty had
submitted a separate appellate brief, arguing that the
conflict between class members that were sued by AIG
and those that weren’t required division of the class
into subclasses, each with separate counsel. That
was an argument on behalf of class members who
were unrelated to the Liberty group.
The second mistake in the majority opinion is related:
it is the statement that “all members of the class are
large and sophisticated businesses, many with millions
on the line and legal staffs to protect their interests. Even
the smaller insurers receive more than $100,000 from
the settlement, and if the representatives had been able
to negotiate for the $3 billion the class initially sought,
the average return per insurer would have exceeded
$2 million (and about $750,000 apiece for the smaller
insurers).” There are 1363 class members, and 55 percent
of the settlement goes to just four of them, leaving
10 Nos. 12-1157, et al.
$202 million to be divided among the other 1359. That is
an average of only $148,638.87 apiece. It implies that
many of the claims probably are much smaller. There is
no basis for the assertion that all the insurers will
receive at least $100,000 from the settlement, and no basis
for estimating the maximum likely settlement.
Some class members had been countersued by AIG,
but others had not been. The brief filed by Liberty’s
subsidiaries argued that those who had not been had
been undercompensated by the settlement because
there was no reason to offset their claims by AIG coun-
terclaims. Those class members might have relied on
the brief of Liberty’s subsidiaries to advance this
argument, foregoing the expense of filing their own
appeals from the class action and their own appeal
briefs because their own claims may not have been
large enough to justify the expense—and anyway why
incur it when they had a champion, namely Liberty?
And remember that, as far as we know, the class mem-
bers have not been informed of the settlement of the
appeals or of the motion to dismiss them. And so the
appellate settlement may be a device by which AIG
paid Liberty to desert the class on whose behalf (as well
as its own) it purported to be appealing.
Rule 42(b) of the appellate rules does not require dis-
missal if the rule’s conditions for dismissal are satisfied;
it says the court “may” dismiss if they are. Further
process is necessary in this case before dismissal can
be considered the responsible course for us to take. The
class action device, as a substitute for individual suits
or conventional joinder, can achieve economies in multi-
Nos. 12-1157, et al. 11
party litigation and allow victims of wrongful acts to
obtain legal relief they couldn’t otherwise obtain. But
class actions are also rife with distorted incentives and
conflicts of interest, which makes judicial review of
class action settlements, whether at the trial or the ap-
pellate level, vital.
This class action suit charged AIG with having
cheated other companies that write workers’ compensa-
tion insurance (the class members) and are required by
state statutes to contribute to a workers’ compensation
insurance liability pool (analogous to an assigned-risk
pool for automobile liability insurance). Employers who
cannot find an insurer willing to write them a workers’
compensation insurance policy because their business
involves a high risk of injury to their employees ob-
tain insurance from the pool.
Allocation of the cost of the liability pool among its
members is based on the amount of workers’ compensation
insurance that each member writes willingly. The suit
charged that AIG cheated the other members of the pool by
underreporting the premiums it received for the
workers’ compensation insurance that it wrote will-
ingly. Its underreporting is alleged to have caused the
class members to pay a portion of what should have
been AIG’s contribution to the pool. Liberty, one of the
class members, became the named plaintiff and sought to
become the class representative. Actually it designated
two of its subsidiaries to be the named plaintiffs—the
two subsidiaries that filed the separate appellate brief
that I mentioned.
12 Nos. 12-1157, et al.
AIG responded by filing suits against a number of
the members of the class, including Liberty, charging
that they were cheaters too, because they too had
underreported their premiums from the insurance
they sold willingly. Liberty responded by filing coun-
terclaims against AIG in AIG’s suit against it. The counter-
claims accused AIG of having underreported premiums
in a number of states not involved in the class action.
Liberty is the only member of the class that has
individual as well as class claims against AIG.
AIG agreed to pay $450 million to the class to settle
both the class claims and Liberty’s individual claims.
Liberty wanted more. It argued that its counterclaims
gave it leverage over AIG that should make AIG agree
to a more generous settlement, because a settlement
would buy AIG peace in the form of a release of those
counterclaims. Liberty wanted to be compensated for
selling AIG that peace. But it didn’t want just a bigger
share of $450 million. It argued that AIG’s offer to the
class was far too low, and not only because of the value
Liberty assigned to its counterclaims. It wanted AIG
to agree to pay $3.1 billion in settlement of the class
action, of which $700 million would go to Liberty
instead of a mere $99 million, its share of the $450
million ultimately awarded in the settlement. Of course
if AIG could be forced to pay $3.1 billion, all the class
members—not just Liberty—would be better off. But
AIG was unwilling.
With Liberty holding up settlement by its intransigence,
ACE INA Holdings, Inc. and six other class members
intervened in the district court, becoming parties. They
Nos. 12-1157, et al. 13
were appointed class representatives for a settlement
class, accepted AIG’s $450 million settlement offer, and
asked the district judge to approve the settlement,
which he did.
Only one class member objected to certification of the
settlement class and ultimately to the settlement it-
self—Liberty. Its objections, renewed in these appeals
that the majority has decided to dismiss blind, were not
only that the settlement was too small but also, as I men-
tioned earlier, that the allocation of the $450 million
among class members ignored the fact that some of them
had been targets of counterclaims by AIG. Their share of
the settlement should have been offset to reflect the value
to them of AIG’s releasing those counterclaims, while the
share received by the class members who had not been
targets of AIG’s counterclaims should have been corre-
spondingly increased—but were not. Instead the settle-
ment money was divided in proportion to each class
member’s share of the liabilities that it had incurred as a
result of AIG’s misconduct, without any offsets. That is the
basis of the argument advanced by Liberty’s subsidiaries
in their separate brief that the judge should have created
two subclasses with separate counsel, one for the members
who were named in AIG’s counterclaims (and thus bene-
fited from the release of those counterclaims, which was
part of the settlement) and the other for those class mem-
bers who weren’t. Representation of a class by one plaintiff
or one group of plaintiffs is inadequate under Fed. R. Civ.
P. 23(a)(4), (g)(4) if there is a potential dispute between
factions within the class over allocation of settlement
proceeds. Ortiz v. Fibreboard Corp., 527 U.S. 815, 856-59
14 Nos. 12-1157, et al.
(1999); Amchem Products, Inc. v. Windsor, 521 U.S. 591, 625-
28 and n. 20 (1997); In re Literary Works in Electronic Data-
bases Copyright Litigation, 654 F.3d 242, 249-53 (2d
Cir. 2011).
Liberty’s unique individual claim to have been under-
paid in the settlement because it was forced to release
its counterclaims against AIG too cheaply has been re-
solved by the settlement with AIG of Liberty’s appeal.
The money for that settlement—the money AIG is
paying to persuade Liberty to drop its appeal—will not
come out of the $450 million of class settlement
money. Were it the only claim, therefore, summary dis-
missal of Liberty’s appeal under Fed. R. App. P. 42(b)
would be proper. But since it’s not the only claim, to
allow Liberty's subsidiaries to withdraw their objection to
the size of the settlement and to the alleged misallocation
of settlement proceeds among the remaining class mem-
bers could deny the class a shot at a larger and more
equitably distributed settlement. If, pursuant to Liberty’s
submission, AIG paid $3.1 billion in settlement, of which
$700 million went to Liberty, $2.4 billion would go to
the rest of the class rather than the $351 million
($450 million minus $99 million) that it will receive if
the settlement approved by the district court stands.
As amended in 2003, Fed. R. Civ. P. 23(e) (in what is
now subsection (e)(5)) says that “an objection [to a class
action settlement] may be withdrawn only with the
court’s approval.” As the committee note points out, the
logic of the rule applies to the withdrawal of an objec-
tion on appeal. “Once an objector appeals, control of the
Nos. 12-1157, et al. 15
proceeding lies in the court of appeals. The court of
appeals may undertake review and approval of a settle-
ment with the objector, perhaps as part of appeal settle-
ment procedures, or may remand to the district court
to take advantage of the district court’s familiarity with
the action and settlement.” 2003 Committee Notes to
Fed. R. Civ. P. 23(e). My concern is that the opposition
of Liberty’s subsidiaries to the settlement may have led
other members of the class not to appeal the alloca-
tion of the settlement proceeds, trusting that someone
(namely the Liberty group) was carrying that ball for
them. Class counsel, it is true, is not objecting to the
dropping of the appeal. But if class counsel could
always be trusted to be the loyal and competent repre-
sentative of the class, there would be no requirement
that class action settlements be submitted for approval
by a court, with approval dependent on the outcome of
a hearing to determine the fairness of the settlement to
the class. “We and other courts have often remarked the
incentive of class counsel, in complicity with the defen-
dant’s counsel, to sell out the class by agreeing with
the defendant to recommend that the judge approve a
settlement involving a meager recovery for the class
but generous compensation for the lawyers—the deal
that promotes the self-interest of both class counsel and
the defendant and is therefore optimal from the stand-
point of their private interests.” Creative Montessori
Learning Centers v. Ashford Gear LLC, 662 F.3d 913, 918
(7th Cir. 2011).
For all we know, the amount that AIG has agreed
to pay Liberty to drop its appeal is not just an estimate
16 Nos. 12-1157, et al.
of the value of Liberty’s individual claim beyond its
share of the class action settlement, but includes a “bribe”
given to Liberty by AIG to take the issue of equitable
allocation of settlement proceeds among class members
out of contention because the issue if taken up by the
appellate court (by us, that is) might be resolved
against approving the settlement. I have pointed out
that members of the class who are disappointed by the
existing allocation may have been counting on Liberty
to champion their cause in this court. But class action
settlements require judicial review (the “fairness” hearing)
even when there are no objectors, in recognition of the
conflicts of interest that pervade class action litigation.
Fed. R. Civ. P. 23(e), (e)(2); 4 William B. Rubenstein et al.,
Newberg on Class Actions §11:48 (4th ed. 2012) (“despite
a lack of opposition, the court should not lose sight of
its responsibility to analyze independently and intelli-
gently the settlement”); Federal Judicial Center, Manual
for Complex Litigation § 21.61 (4th ed. 2004); cf. Mirfasihi
v. Fleet Mortgage Corp., 551 F.3d 682, 686-87 (7th Cir.
2008); In re General Motors Corp. Pick-Up Truck Fuel
Tank Products Liability Litigation, 55 F.3d 768, 812-13
(3d Cir. 1995).
So how should we proceed? We could remand the case
to the district court for a determination of whether
to approve the dismissal of Liberty’s appeal. But that
would inject needless delay. A superior alternative
would be to conduct our own investigation of whether
to approve the settlement between Liberty and AIG. The
first step would be simply to require submission to us
of the settlement agreement. Maybe on reading it we’d
Nos. 12-1157, et al. 17
conclude that it is innocuous and dismiss the appeals.
But maybe not. According to Liberty its independent (non-
class) claim against AIG was valued at $25 million in
the district court settlement. If Liberty’s appellate settle-
ment with AIG exceeds that amount, this may be a
clue that AIG is paying Liberty to drop objections to
the settlement that, were they accepted, would benefit
the class. In that event the class is being hurt by the
blind withdrawal of Liberty’s appeal unless no more
money can be squeezed out of AIG, which we don’t know.
We should not dismiss the appeal without at least in-
forming ourselves of the terms of Liberty’s settlement
with AIG. In dismissing the appeals without doing so
we are acting in haste, and for no good reason. The
motion to dismiss the appeals was filed more than two
months ago. Rather than arguing over whether to
dismiss them we could within this period have com-
pleted the investigation that would reveal whether
we should grant the motion.
3-25-13