In the
United States Court of Appeals
For the Seventh Circuit

No. 00-1407

Sybron Transition Corporation and Kerr
Manufacturing Corporation,

Plaintiffs-Appellants,

v.

Security Insurance of Hartford,

Defendant-Appellee.

Appeal from the United States District Court
for the Eastern District of Wisconsin.
No. 92-C-779--Lynn Adelman, Judge.

Argued October 27, 2000--Decided July 12, 2001



  Before Easterbrook, Kanne, and Rovner,
Circuit Judges.

  Easterbrook, Circuit Judge. This appeal
concerns insurance for asbestos
liabilities. The estate of Alan Press, a
dentist who died in 1988 of mesothelioma,
contended that his disease had been
caused by exposure to asbestos during
dental school from September 1969 through
May 1973, and that the source of the
asbestos was products made by Kerr
Manufacturing Corporation. That suit was
settled for $1.3 million, of which
Security Insurance contributed $500,000
under a reservation of rights. Kerr and
its parent Sybron Transition Corp.
(collectively Sybron) contend in this
suit under the diversity jurisdiction
that Security must indemnify them for the
entire settlement; Security replies that
it is entitled to most of its $500,000
back. After a bench trial the district
judge concluded that Security’s share of
the liability is $230,208, and he entered
a judgment requiring Sybron to refund the
excess. 2000 U.S. Dist. Lexis 19139 (E.D.
Wis. Jan. 14, 2000).
  Security underwrote Sybron’s tort
liability when Press arrived at dental
school; its last policy expired at the
end of January 1971, during his
education. Mesothelioma was diagnosed in
July 1987, Press died in April 1988, and
his estate filed suit in 1989. By 1986
Sybron was self-insuring for most
asbestos risks. One possible resolution
of coverage disputes would have been to
say that the responsibility for indemnity
fell on Sybron’s carrier in 1984, when
(the district judge found) Press’s tumor
began growing, or on Sybron itself if
1987 marked the onset of the disease. But
the parties agree that New York law,
which governs the application of
Security’s policy, applies a time-on-the-
risk approach to allocating insurance
coverage for diseases with long latency
periods (and similar matters such as
pollution)--and that New York does this
essentially no matter how the insurance
policy defines the conditions of its own
coverage. The parties have accordingly
paid little attention to the language of
the policies and a great deal of
attention to Stonewall Insurance Co. v.
Asbestos Claims Management Corp., 

(2d Cir. 1995), and Olin Corp. v.
Insurance Co. of North America, 

(2d Cir. 2000). These are the leading
decisions about this aspect of New York
insurance law--true, they are not
decisions by New York courts, but the
parties treat them as authentic
expositions of New York law. So if in a
usual diversity case the federal court
acts as a ventriloquist’s dummy for the
state judiciary, we are playing this hand
double dummy: the second circuit has
interpreted New York law, and we are
interpreting the work of the second
circuit.

  Time-on-the-risk means that each
insurer’s liability (up to its policy
limit) is measured by the underlying loss
multiplied by the ratio of time covered
by the policy to the time subject to the
risk. The denominator of this fraction,
the total period of risk, was set by the
district court after the bench trial at
96 months: the 45 months Press was in
dental school (and exposed to asbestos)
plus the 51 months during which cancerous
cells were multiplying in Press’s body.
The numerator, according to the district
court, is 17 months: the portion of
Press’s dental education during which
Sybron had coverage from Security.
Multiplying $1.3 million by the fraction
17/96 produced Security’s share
($230,208), less than the $500,000 per-
occurrence limit in Security’s policies.
Sybron contends that the district judge
made three errors in working this out.
Sybron contends first that the numerator
should be 36 months rather than 17;
second that the denominator should be 69
months rather than 96; and third that
Security’s three policies should be
stacked to increase its maximum exposure
to $1.5 million.

  1. The numerator. Security wrote three
policies that covered portions of the
risk: one for calendar year 1969, a
second for calendar year 1970, and a
third for the month of January 1971,
bridging a gap while Sybron arranged for
coverage from another underwriter. Sybron
contends that because three policies are
at issue, it is entitled to have Security
treated as covering three years of the
risk. The year is the ordinary unit of
insurance coverage, Sybron observes, so
it should be the unit of allocation for
time-on-the-risk calculations too.

  A year is the normal accounting period
for financial reports and tax returns,
but whether it is the right accounting
period for insurance is something that
parties can and do work out for
themselves. Security wrote a one-month
policy for 1971 (at Sybron’s request) and
presumably charged one-twelfth of its
annual premium. Yet Sybron believes that
it obtained the same coverage as a
premium twelve times larger would have
produced. That does not seem sensible and
is not supported by anything in the
policy’s language. If one of Kerr’s
products had killed Press outright
inJanuary 1971 then Security would have
been responsible for up to $500,000; but
a death in February 1971 would have been
the responsibility of Security’s
successor. If liability for asbestos-
related disease were allocated to a
single policy-- say, the policy in force
at first exposure, or the policy in force
when the disease becomes manifest--again
the single-month policy for January 1971
would subject Security to one month’s
actuarial risk. A twelfth of all initial
exposures that happened anytime in a year
would come in January, and a twelfth of
all mesotheliomas would be discovered
that month.

  The same approach applies to a time-on-
the-risk calculation: if the events are
to be allocated among the underwriters
according to total exposure (and
manifestation) time, then a one-month
policy covers one month’s worth of
exposure rather than a year’s worth. Both

, 1217, and

, support this
conclusion. True, both opinions
frequently refer to "years" in the
numerator, but that is because they dealt
with whole-year policies. When insured
and insurer have elected to parcel out
coverage by month rather than by year,
the formula should use months rather than
years. (And, we suppose, if coverage were
measured in days, then days would be the
right unit for the formula. Cf. National
Casualty Insurance Co. v. Mt. Vernon, 

, 

(App. Div.
1987).) Otherwise the insured would
receive coverage it did not pay for. That
the underwriter covering the remainder of
1971 (American Mutual) became insolvent
does not increase Security’s liability;
Sybron does not point to any provision of
New York law making an insurer
responsible for coverage its successor
promised but failed to deliver.

  2. The denominator. The district court
included in the denominator all of the
months that Press either was in dental
school or suffered from cancer. Sybron
contends that the denominator should
include only the period during which it
carried insurance that would have covered
the loss. Sybron was not wholly self-
insured; it carried insurance for risks
exceeding $2 million per occurrence; but
this coverage did not kick in until after
the liability level of the Press
litigation. Both Sybron and the district
court call Sybron’s strategy "self-
insurance" and we will follow suit,
though it would be more accurate to call
it "insurance with a big deductible."

  Sybron concedes that self-insurance is
a form of insurance but contends that
time during which it self-insured out of
necessity should be excluded from the
denominator. This legal proposition has
the support of both 

, and 

. The
district court followed this approach but
concluded that, even if asbestos coverage
was unavailable (a subject on which the
court reserved decision), this is
irrelevant because Sybron had decided to
self-insure come what may. To this Sybron
responds that Stonewall determines that
asbestos coverage is "unavailable" for
its purposes whenever it cannot be
obtained as an ordinary part of a
comprehensive general liability policy--
and no one denies that by 1986
comprehensive general liability policies
excluded injuries caused by exposure to
asbestos. Thus it is entitled to have
time after 1985 removed from the
denominator no matter what other
insurance may have been available, and no
matter what its plans may have been.

  Security was not a party to Stonewall,
so we do not see how the second circuit’s
views about the availability of asbestos
coverage can be conclusive against it.
All Stonewall could do--all we can do--is
predict how the courts of New York would
handle a legal issue that has never been
presented to them. We are not confident
that Stonewall itself held that only
comprehensive general coverage counts as
"available" coverage for the purposes of
New York law. That was not a subject
debated by the parties or addressed
squarely by Stonewall. Indeed, we do not
know what it means (or could mean) to say
that coverage for a particular risk is
"unavailable." Unavailable at what price?

  Underwriters cheerfully sell insurance
policies even after a risk has come to
pass and the obligation to pay is
certain. What they then deliver on such
retroactive policies is a claims-
administration service rather than risk-
spreading, and the premium exceeds the
expected amount of the outlay (for the
underwriter must be reimbursed for both
the expected payments to victims and the
expected costs of evaluating and
resolving claims). This kind of insurance
could have been purchased for asbestos
risks in the mid-1980s, though perhaps no
less expensive policy would have been
available. For by the 1980s it had become
clear that firms such as Sybron that used
asbestos in their business had
accumulated large potential liabilities,
and any underwriter of asbestos risks
faced substantial liability for diseases
that became manifest during the policy
period, even if the insured had long
since stopped using that mineral. A
policy covering asbestos liability
therefore had become, by the mid-1980s, a
form of retrospective coverage of
diseases with long latency periods. It
was impossible to spread risks: the
casualties had occurred and only
manifestation remained. No firm outside
those industries that used asbestos would
purchase asbestos coverage; the only
firms that wanted this expensive coverage
already had a backlog of dormant
liability, so adverse selection would
confine interest in asbestos coverage to
the firms with the greatest overhang of
liability for acts in years past. That
would lead insurers to price the coverage
at the full policy limits, plus loading
charges and administration fees. And
firms such as Sybron, whose asbestos-
related liability was below that of mine
operators, shipyards, and construction
firms, would not find attractive a
premium that exceeded the policy limits.
Knowing that its own anticipated
liability was smaller, Sybron opted out
of the pool and self-insured. This is ad
verse selection at work, an ordinary
phenomenon in insurance markets when
would-be insureds know more about their
risks than do insurers. Instead of saying
either that insurance was "available" or
"unavailable," or that Sybron was bound
and determined to self-insure, it is
better to say that Sybron did not in the
late 1980s have an economically
attractive opportunity to participate in
a pool in which the risks of asbestos-
related casualties were spread among sim
ilar firms.

  What, then, is the consequence under New
York law of a low-risk firm opting to
self-insure because the alternative is a
premium structure tailored to high-risk
members of the pool? Stonewall and Olin
are of little help on that score, because
they assume that insurance is "available"
or "unavailable," as if it were on or off
like a light bulb rather than a commodity
whose price may be attractive or
unattractive to a given customer. No
decision by a state court in New York
addresses the question. We have to tackle
it from scratch, and from that
perspective the answer is clear: self-
insurance is an alternative to market
insurance. Sybron "paid" (to itself) a
"premium" for coverage that was less than
the one commercial insurers sought to
charge, and it then bore the ensuing
risks. But instead of asking whether
Sybron had some kind of insurance, we
prefer to ask why it should matter
whether Sybron was insured. Suppose it
had not set up any casualty reserves for
1986-88 (the equivalent of the coverage
"purchased" with the internal "premium"
that Sybron saved and released to other
uses). Why would this affect the legal
obligations of a firm whose last policy
expired in 1971? The whole idea of a
time-on-the-risk calculation is that any
given insurer’s share reflects the ratio
of its coverage (and thus the premiums it
collected) to the total risk. The full
risk is not affected by whether insurance
is available later.

  To say that developments in the 1980s
increased the exposure of an underwriter
such as Security would be equivalent to
saying that the only relevant "risk" is
measured by the period of exposure.
(Recall that the reason asbestos coverage
was so expensive in the late 1980s was
the manifestation of diseases caused by
exposures long ago.) Yet no one argues
for that proposition. Security wrote an
occurrence policy; and for occupational
diseases the time of the "occurrence" (or
the "accident") traditionally had been
understood as the time the disease
becomes manifest. That might justify
assigning the whole Press loss to
Sybron’s carriers during 1984-88. The
difficulty of assigning causation might
justify splitting responsibility among
the periods of exposure and
manifestation. See E.R. Squibb & Sons,
Inc. v. Lloyd’s & Cos., 

(2d
Cir. 2001); Fogel v. Zell, 

(7th Cir. 2000); Eljer Manufacturing,
Inc. v. Liberty Mutual Insurance Co., 

(7th Cir. 1992). But we do not
know of any support in New York law for
assigning all (effective) coverage to the
carriers during the period of exposure--
and certainly not for concentrating
responsibility on an early carrier just
because coverage was too expensive later.

  To require Security to pay extra because
Sybron did not find it cost-effective to
purchase coverage during 1986 to 1988
would be the economic equivalent of
requiring Security to furnish free
coverage during 1986-88 (for Sybron does
not propose to pay the going premium
retroactively). Why an underwriter who
furnishes low-price coverage during a
period before the magnitude of the risk
became apparent should be required to
furnish, for nothing, an additional
period of high-price coverage escapes us.
After all, it was Sybron, not Security,
that created the risk of loss. And the
consequences of that risk should fall on
its creator, not on an underwriter
unlucky enough to insure an early slice
of the risk. Sybron could choose to
handle risks as it pleased. It is a
little unclear why corporations buy
insurance at all (for their shareholders
can diversify risks more readily in the
stock market than through insurance),
making it doubly obscure why a firm’s
rational decision to depend on
shareholders for riskbearing services in
the band under $2 million should cause a
former insurer to pony up.

  3. Stacking (aka joint and several
liability). Hoping to avoid all of this
time slicing, Sybron argues that it can
pick any policy during the period of risk
and require its insurer to bear the whole
loss up to the limit of liability (which
in Security’s case was $500,000 per
occurrence). If this does not cover the
loss, the insured names a second policy,
and then a third. Sybron hopes in this
manner to make Security pick up the
entire tab. It would require Security to
bear $500,000 on the 1969 policy, another
$500,000 on the 1970 policy, and the
remaining $300,000 on the January 1971
policy. The strategy is known variously
as "stacking" and "joint and several
liability"--though this use of the latter
phrase is unusual. See 

. See also Comment, Allocating
Progressive Injury Liability Among
Successive Insurance Policies, 64 U. Chi.
L. Rev. 257 (1997).

  No matter what the right name of this
possibility, it is antithetical to a
time-on-the-risk approach. Courts have
adopted the time-on-the-risk method
because it is impossible to tell whether
fibers of asbestos inhaled in a given
year caused any given asbestos-related
disease. Granted, mesothelioma (unlike
asbestosis) does not depend on cumulative
exposure. Brief exposure may suffice. But
when did the fatal exposure occur? In
1969, or 1970, or maybe in 1973 when
Security was not the insurer, or maybe
even when Press was not at school; all
are possible. Instead of trying to pursue
this causal will-o’-the-wisp, courts
allocate liability to all periods that
are arguably appropriate--to periods of
exposure because they contain the likely
causes, to periods of manifestation
because they contain the consequences
(and are most closely associated with the
injury that would mark the "occurrence"
or "accident" triggering coverage in a
standard policy). Liability is spread
among insurers to reflect the uncertainty
in the timing of cause and consequence.
  Sybron wants to combine this
uncertainty-based approach, which defines
a range of eligible policies, with an
entitlement to choose a particular policy
for indemnity--yet collecting all of the
indemnity from a particular policy
supposes ability to pin down the cause.
Sybron does not propose to demonstrate
that asbestos fibers inhaled during 1969
caused Press’s mesothelioma, so it has no
basis for insisting that the whole policy
limit for 1969 be made available.
(Actually, Sybron tells us, the kind of
asbestos used in its products does not
cause mesothelioma. By settling Press’s
suit Sybron gave up any chance at putting
this defense across, but this line of
argument makes it doubly hard to see how
Sybron can turn around and express
confidence that 1969, or for that matter
1970 or 1971, was the causal period.)
Olin explains in 

-
24, why a time-on-the-risk allocation is
superior to allowing insureds to pick any
(or every) policy during the risk period.
No New York court has addressed this
question, see Continental Casualty Co. v.
Rapid-American Corp., 

, 

(1993) (reserving the point),
and we think that Olin has anticipated
the answer New York will give when the
time arrives.

  Sybron believes that the language of
these particular policies is on its side,
but to the contrary the language strongly
suggests no coverage at all. Security
promised to indemnify Sybron for all sums
that Sybron becomes legally obliged to
pay as a result of bodily injury that is
"caused by an occurrence". A separate
clause defines "occurrence" this way: "an
accident, including injurious exposure to
conditions which results during the
policy period in bodily injury." Thus the
"bodily injury" must occur "during the
policy period". One way to read this
would be to say that until the symptoms
of the disease are manifest there is no
"bodily injury"; then Security’s policies
would not cover any of the Press loss.
The other way to read it would be to say
that the "bodily injury" occurs when the
asbestos fibers pierce the lung wall and
lodge in the pleura, beginning the
process that leads to mesothelioma. But
this drives us back to the causation
question. Did this occur in 1969, 1970,
January 1971, or some other time? No one
knows. The only way to solve, or at least
mitigate, the causation problem is to
adopt a time-on-the-risk approach, which
Olin sensibly holds is incompatible with
allowing the insured to pick and choose
among policies.

  What we have to add to Olin is that even
if knowledge of causation permits an
insured to pick a policy, it may not pick
more than one. Stacking is incompatible
with confidence about causation. Security
insured Sybron to a limit of $500,000 per
occurrence, not $500,000 per occurrence
per year. Suppose Press had not only
inhaled asbestos but also developed
mesothelioma in 1969 as a result. How
much could Sybron have recovered? Surely
the answer is $500,000 maximum; the
occurrence or accident would be confined
to 1969, and the fact that Security wrote
another policy the next year would not
justify treating one casualty as multiple
occurrences just because the victim lived
into 1970. This would be clear enough for
an auto accident that caused medical
expenses and lost income not only in the
year of the collision but also in future
years; it is no less true of occupational
diseases. There is only one "occurrence"
no matter how many years the loss
extends. A time-on-the-risk approach
spreads responsibility among insurers to
reflect uncertainties about causation,
but it does not justify treating one loss
as more than one occurrence and requiring
insurers individually (or in the
aggregate) to pay more than the
occurrence limit of their policies.

Affirmed