Reinsurance

Chapter 7
REINSURANCE
GARY S. PATRIK
INTRODUCTION
What is Reinsurance?
Reinsurance is a form of insurance. A reinsurance contract is
legally an insurance contract. The reinsurer agrees to indemnify
the cedant insurer for a specified share of specified types of in-
surance claims paid by the cedant for a single insurance policy
or for a specified set of policies. The terminology used is that
the reinsurer assumes the liability ceded on the subject policies.
The cession, or share of claims to be paid by the reinsurer, may
be defined on a proportional share basis (a specified percentage
of each claim) or on an excess basis (the part of each claim, or
aggregation of claims, above some specified dollar amount).
The nature and purpose of insurance is to reduce the finan-
cial cost to individuals, corporations, and other entities arising
from the potential occurrence of specified contingent events. An
insurance company sells insurance policies guarantying that the
insurer will indemnify the policyholders for part of the financial
losses stemming from these contingent events. The pooling of
liabilities by the insurer makes the total losses more predictable
than is the case for each individual insured, thereby reducing
the risk relative to the whole. Insurance enables individuals, cor-
porations and other entities to perform riskier operations. This
increases innovation, competition, and efficiency in a capitalistic
marketplace.
The nature and purpose of reinsurance is to reduce the fi-
nancial cost to insurance companies arising from the potential
occurrence of specified insurance claims, thus further enhancing
innovation, competition, and efficiency in the marketplace. The
cession of shares of liability spreads risk further throughout the
343
344 REINSURANCE Ch. 7
insurance system. Just as an individual or company purchases an

insurance policy from an insurer, an insurance company may pur-
chase fairly comprehensive reinsurance from one or more rein-
surers. A reinsurer may also reduce its assumed reinsurance risk
by purchasing reinsurance coverage from other reinsurers, both
domestic and international; such a cession is called a retroces-
sion.
Reinsurance companies are of two basic types: direct writers,
which have their own employed account executives who produce
business, and broker companies or brokers, which receive busi-
ness through reinsurance intermediaries. Some direct writers do
receive a part of their business through brokers, and likewise,
some broker reinsurers assume some business directly from the
ceding companies. It is estimated that more than half of U.S.
reinsurance is placed via intermediaries.
The form and wording of reinsurance contracts are not as
closely regulated as are insurance contracts, and there is no rate
regulation of reinsurance between private companies. A rein-
surance contract is often a manuscript contract setting forth the
unique agreement between the two parties. Because of the many
special cases and exceptions, it is difficult to make correct gen-
eralizations about reinsurance. Consequently, as you read this
chapter, you should often supply for yourself the phrases “It is
generally true that: : :” and “Usually: : :” whenever they are not
explicitly stated.
This heterogeneity of contract wordings also means that
whenever you are accumulating, analyzing, and comparing var-
ious reinsurance data, you must be careful that the reinsurance
coverages producing the data are reasonably similar. We will be
encountering this problem throughout this chapter.
The Functions of Reinsurance

Reinsurance does not change the basic nature of an insurance
coverage. On a long-term basis, it cannot be expected to make
INTRODUCTION 345
bad business good. But it does provide the following direct as-
sistance to the cedant.
Capacity
Having reinsurance coverage, a cedant can write higher pol-
icy limits while maintaining a manageable risk level. By ceding
shares of all policies or just larger policies, the net retained loss
exposure per individual policy or in total can be kept in line with
the cedant’s surplus. Thus smaller insurers can compete with
larger insurers, and policies beyond the capacity of any single
insurer can be written.
The word “capacity” is sometimes also used in relation to ag-
gregate volume of business. This aspect of capacity is best con-
sidered below in the general category of financial results man-
agement.
Stabilization
Reinsurance can help stabilize the cedant’s underwriting and
financial results over time and help protect the cedant’s sur-
plus against shocks from large, unpredictable losses. Reinsur-
ance is usually written so that the cedant retains the smaller,
predictable claims, but shares the larger, infrequent claims. It
can also be written to provide protection against a larger than
predicted accumulation of claims, either from one catastrophic
event or from many. Thus the underwriting and financial effects
of large claims or large accumulations of claims can be spread
out over many years. This decreases the cedant’s probability of
financial ruin.
Financial Results Management

Reinsurance can alter the timing of income, enhance statutory
and/or GAAP surplus, and improve various financial ratios by
which insurers are judged. An insurance company with a grow-
ing book of business whose growth is stressing their surplus can
cede part of their liability to a reinsurer to make use of the rein-
surer’s surplus. This is essentially a loan of surplus from the
reinsurer to the cedant until the cedant’s surplus is large enough

to support the new business. We will see other ways that rein-
surance can be used to alter a cedant’s financial numbers. As
you might expect in a free market, this aspect of reinsurance
has led to some abuses in its use. As we discuss the various
forms of reinsurance coverage, we will note their financial ef-
fects.
Management Advice
Many professional reinsurers have the knowledge and ability
to provide an informal consulting service for their cedants. This
service can include advice and assistance on underwriting, mar-
keting, pricing, loss prevention, claims handling, reserving, actu-
arial, investment, and personnel issues. Enlightened self-interest
induces the reinsurer to critically review the cedant’s operation,
and thus be in a position to offer advice. The reinsurer typically
has more experience in the pricing of high limits policies and
in the handling of large and rare claims. Also, through contact
with many similar cedant companies, the reinsurer may be able
to provide an overview of general issues and trends. Reinsurance
intermediaries may also provide some of these same services for
their clients.
The Forms of Reinsurance

Facultative Certificates
A facultative certificate reinsures just one primary policy. Its
main function is to provide additional capacity. It is used to
cover part of specified large, especially hazardous or unusual
exposures to limit their potential impact upon the cedant’s net
results or to protect the cedant’s ongoing ceded treaty results in
order to keep treaty costs down. The reinsurer underwrites and
accepts each certificate individually; the situation is very simi-
lar to primary insurance individual risk underwriting. Because
facultative reinsurance usually covers the more hazardous or un-
usual exposures, the reinsurer must be aware of the potential for
antiselection within and among classes of insureds.
INTRODUCTION 347
Property certificate coverage is sometimes written on a pro-

portional basis; the reinsurer reimburses a fixed percentage of
each claim on the subject policy. Most casualty certificate cover-
age is written on an excess basis; the reinsurer reimburses a share
(up to some specified dollar limit) of the part of each claim on
the subject policy that lies above some fixed dollar attachment
point (net retention).
Facultative Automatic Agreements or Programs

A facultative automatic agreement reinsures many primary
policies of a specified type. These policies are usually very sim-
ilar, so the exposure is very homogeneous. Its main function is
to provide additional capacity, but since it covers many policies,
it also provides some degree of stabilization. It may be thought
of as a collection of facultative certificates underwritten simulta-
neously. It may cover on either a proportional or excess basis. It
is usually written to cover new or special programs marketed by
the cedant, and the reinsurer may work closely with the cedant
to design the primary underwriting and pricing guidelines. For
example, a facultative automatic agreement may cover a 90%
share of the cedant’s personal umbrella business, in which case
the reinsurer will almost certainly provide expert advice and will
monitor the cedant’s underwriting and pricing very closely.
Facultative automatic agreements are usually written on a
fixed cost basis, without the retrospective premium adjustments
or variable ceding commissions sometimes used for treaties (as
we shall see below).
There are also non-obligatory agreements where either the
cedant may not be required to cede or the reinsurer may not
be required to assume every single policy of the specified type.
Treaties
A treaty reinsures a specified part of the loss exposure for
a set of insurance policies for a specified coverage period. For
ongoing treaty coverage, the claims covered may be either those
occurring during the treaty term or those occurring on policies

written during the term. In the case of claims-made coverage, the
word “occurring” means those claims made to the ceding com-
pany during the term. The premium subject to the treaty cor-
responds to the types of claims covered: it is earned premium
arising from policies of the specified type either in force or writ-
ten during the term of the treaty. The subject exposure is usually
defined by Annual Statement line of business or some variant or
subsets thereof. Because an ongoing treaty relationship involves
a close sharing of much of the insurance exposure, it can create a
close working partnership between the parties; the expertise and
services of the reinsurer or broker are available to the cedant.
This is especially true for treaties written by a direct writer or
where there is a strong reinsurer leading a brokered treaty.
Treaty Proportional Covers

A quota-share treaty reinsures a fixed percentage of each sub-
ject policy. Its main function is financial results management,
although it also provides some capacity. The reinsurer usually
receives the same share of premium as claims, and pays the
cedant a ceding commission commensurate with the primary pro-
duction and handling costs (underwriting, claims, etc.). Quota-
share treaties usually assume in-force exposure at inception. The
cedant’s financial results are managed because the ceding com-
mission on the ceded unearned premium reserve transfers statu-
tory surplus from the reinsurer to the cedant. (We shall see this
later.) The cession of premium also reduces the cedant’s net-
premium-to-surplus ratio.
The ceding commission on quota-share treaties is often de-
fined to vary within some range inversely to the loss ratio. This
allows the cedant to retain better-than-expected profits, but pro-
tects the reinsurer somewhat from adverse claims experience.
The term quota-share is sometimes (mis-)used when the cov-
erage is a percentage share of an excess layer; we will more
properly treat this kind of coverage as being excess.
INTRODUCTION 349
A surplus-share treaty also reinsures a fixed percentage of

each subject policy, but the percentage varies by policy accord-
ing to the relationship between the policy limit and the treaty’s
specified net line retention. Its main function is capacity, but
it also provides some stabilization. A surplus-share treaty may
also assume in-force exposure at inception, which together with
a ceding commission provides some management of financial
results. This is typically a property cover; it is rarely used for
casualty business.
Treaty Excess Covers

An excess treaty reinsures, up to a limit, a share of the part of
each claim that is in excess of some specified attachment point
(cedant’s retention). Its main functions are capacity and stabiliza-
tion. An excess treaty typically covers exposure earned during
its term on either a losses-occurring or claims-made basis, but
run-off exposure may be added in. The definition of “subject
loss” is important.
For a per-risk excess treaty, a subject loss is defined to be the
sum of all claims arising from one covered loss event or occur-
rence for a single subject policy. Per-risk excess is mainly used
for property exposures. It often provides protection net of facul-
tative coverage, and sometimes also net of proportional treaties.
It is used for casualty less often than per-occurrence coverage.
For a per-occurrence excess treaty, a subject loss is defined to
be the sum of all claims arising from one covered loss event or
occurrence for all subject policies. Per-occurrence excess is used
for casualty exposures to provide protection all the way up from
working cover layers through clash layers.
A working cover excess treaty reinsures an excess layer for
which claims activity is expected each year. The significant ex-
pected claims frequency creates some stability of the aggregate
reinsured loss. So working covers are often retrospectively rated,
with the final reinsurance premium partially determined by the
treaty’s loss experience.
A higher exposed layer excess treaty attaches above the work-

ing cover(s), but within policy limits. Thus there is direct single-
policy exposure to the treaty.
A clash treaty is a casualty treaty that attaches above all policy
limits. Thus it may be only exposed by:
1. extra-contractual-obligations (i.e., bad faith claims)

2. excess-of-policy-limit damages (an obligation on the part
of the insurer to cover losses above an insurance con-
tract’s stated policy limit)
3. catastrophic workers compensation accidents
4. the “clash” of claims arising from one or more loss
events involving multiple coverages or policies.
Both higher exposed layers and clash are almost always priced
on a fixed cost basis, with no variable commission or additional
premium provision.
Catastrophe Covers
A catastrophe cover is a per-occurrence treaty used for prop-
erty exposure. It is used to protect the net position of the cedant
against the accumulation of claims arising from one or more large
events. It is usually stipulated that two or more insureds must be
involved before coverage attaches. The coverage is typically of
the form of a 90% or 95% share of one or more layers (separate
treaties) in excess of the maximum retention within which the
cedant can comfortably absorb a loss, or for which the cedant
can afford the reinsurance prices.
Aggregate Excess, or Stop Loss Covers

For an aggregate excess treaty, also sometimes called a stop
loss cover, a loss is the accumulation of all subject losses during
a specified time period, usually one year. It usually covers all or
part of the net retention of the cedant and protects net results,
providing very strong stabilization. Claims arising from natural
INTRODUCTION 351
catastrophes are often excluded, or there may be a per-occurrence

maximum limit.
Finite, or Nontraditional, Reinsurance Covers

Over the past few years, there has been a growing use of
reinsurance, especially treaties, whose only or main function is
to manage financial results. The word “finite” means that the
reinsurer’s assumed risk is significantly reduced by various con-
tractual conditions, sometimes called “structure.” Of course, the
reinsurer’s expected margin (expense and profit) is also reduced
to reflect this reduced risk transfer. Sometimes these covers are
structured to induce a favorable tax treatment for the cedant. Of-
ten they are based on the ability of offshore reinsurers to book
claims on a discounted basis in anticipation of the future in-
vestment income that will be earned from the premium income
received before the claims are settled. The reinsurance premium
reflects this discounting, thus giving the cedant a statutory and
GAAP accounting benefit.
There have been cases where the risk transfer was nonexistent
or negligible. In order to stop accounting abuses through reinsur-
ance, the Financial Accounting Standards Board issued FAS 113
in 1992. FAS 113 requires a measurable and significant transfer
of risk before a contract can receive the benefit of reinsurance
accounting. Although the standard is somewhat ambiguous, it
has largely stopped abusive reinsurance practices.
There continues to be debate in the reinsurance community
about a workable distinction between the categories: traditional
and finite reinsurance. Other than the rather ambiguous FAS 113,
there is no clear boundary between traditional reinsurance and fi-
nite reinsurance; there is a continuum of risk transfer possibility
between full risk transfer and no transfer. Virtually any rein-
surance contract can be structured in a way to reduce the risk
transfer and become “finite.” We shall see this in the follow-
ing discussion of typical forms for finite reinsurance. Through-
out this chapter, we assume that any reinsurance contract under
discussion has sufficient risk transfer to pass FAS 113 require-

ments.
The first typical form for finite reinsurance is a financial pro-

portional cover. As noted above, proportional treaties quite often
have a ceding commission that varies inversely with the losses;
this limits the risk transfer. The degree of variation can be in-
creased to further limit the risk transfer. Also, the loss share
may be defined to decrease somewhat if the losses exceed some
maximum. Quite often, these treaties may also have some kind
of funding mechanism, wherein the aggregate limit of cover-
age is based upon the fund (net cash position less the reinsurer’s
margin) together with some remote risk layer. Whatever the risk-
limiting structure, the contract must be checked with the cedant’s
accountants to assure that they will approve the risk transfer for
FAS 113 guidelines.
A loss portfolio transfer is also a very prevalent form for finite

reinsurance. This is a retrospective cover, a cession of part of the
cedant’s loss liabilities as of a specified accounting date. It may
be a cession of the total liability or, more often, a cession of
some aggregate excess layer of the liability. An aggregate excess
cover attaching at the cedant’s carried loss reserve is often called
an adverse development cover. It is clear that a loss portfolio
transfer could be a pure risk non-finite cover. To make the risk
transfer finite, it has an aggregate limit and may have sublimits
for various types of claims, and it is priced to be essentially
a present-value funding of liabilities with a substantial upfront
provisional margin for the reinsurer. Part of this margin will be
paid back to the cedant in the form of a profit commission if the
loss experience is favorable.
A funded aggregate excess cover is, as you might expect, an

aggregate excess treaty in which the premium is high enough to
fund the loss payments except in extraordinary circumstances.
It is analogous to a funded loss portfolio transfer except that it
covers future occurring claims. In addition to financial results
INTRODUCTION 353
management, it may provide strong stabilization of the cedant’s

net results.
A Typical Reinsurance Program

There is no such thing as a typical reinsurance program. Every
insurance company is in a unique situation with regard to loss
exposure, financial solidity, management culture, future plans,
and marketing opportunities. Thus each company needs a unique
reinsurance program, a combination of ceded reinsurance covers
tailor-made for that company.
Nevertheless, Table 7.1 displays what we might regard as
a “typical” reinsurance program for a medium-sized insurance
company.
If the company writes surety, fidelity, marine, medical mal-
practice, or other special business, other similar reinsurance cov-
ers would be purchased. If the company were entering a new
market (e.g., a new territory or a new type of business), it might
purchase a quota-share treaty to lessen the risk of the new busi-
ness and the financial impact of the new premium volume, and to
obtain the reinsurer’s assistance. Or it might purchase a propor-
tional facultative automatic agreement for an even closer work-
ing relationship with a reinsurer. If the company were exiting a
market, it might purchase a loss portfolio transfer, especially an
adverse development cover, to cover part of the run-off claims
payments.
The Cost of Reinsurance to the Cedant

The Reinsurer’s Margin
In pricing a reinsurance cover, the reinsurer charges a margin
over and above the ceded loss expectation, commission, and bro-
kerage fee (if any). The margin is usually stated as a percentage
of the reinsurance premium. It is theoretically based upon the
reinsurer’s expenses, the degree of risk transfer, and the magni-
TABLE 7.1
A REINSURANCE PROGRAM
FOR A MEDIUM-SIZED INSURANCE COMPANY
Line of Business Type of Reinsurance
Fire and Allied Lines 1. Proportional and excess facultative certificates to

HO Section I bring each individual policy’s net exposure down to
SMP Section II $2M
2. Surplus share of four lines not to exceed $1.6M;
maximum cedant retention of $400,000
3. Per-risk excess working cover of $200k excess of
$200k
4. Catastrophe covers:
4.1. 95% of $5M excess of $5M
4.2. 95% of $10M excess of $10M
4.3. 95% of $10M excess of $20M
4.4. 95% of $10M excess of $30M
Casualty Lines excluding 1. Facultative certificates for primary per policy
Umbrella coverage excess of $2M
2. Working cover $500k excess of $500k
3. Higher exposed layer $1M excess of $1M
4. Clash layers:
4.1. $3M excess of $2M
Personal Umbrellas 1. 90% share facultative automatic program
tude of capacity and financial support, but it is practically influ-

enced by competition in the reinsurance market. Of course, as
with most insurance, the actual resulting margin will differ from
that anticipated because of the stochasticity of the loss liability
and cash flow transferred.
Brokerage Fee
A reinsurance broker charges a brokerage fee for placing the
reinsurance coverage and for any other services performed on
behalf of the cedant. This fee is incorporated into the reinsur-
INTRODUCTION 355
ance premium and is paid by the reinsurer. Offsetting this cost

is the fact that broker reinsurers usually have lower internal ex-
penses because they don’t maintain separate marketing staffs.
The brokerage fee is usually a fixed percentage of the reinsur-
ance premium, but on occasion may be defined as either a fixed
dollar or as some other variable amount.
Lost Investment Income

For most reinsurance contracts, the premium funds (net of
ceding commission) are paid to the broker, if any, who then
passes them on (also net of brokerage fee) to the reinsurer. The
cedant thus loses the use of those funds, and the reinsurer gains
the investment income earned on those funds until returned as
loss payments, ceding commission adjustments or other premium
adjustments. The price of the reinsurance cover accounts for this
investment income.
Some contracts incorporate a funds withheld provision, where
the cedant pays only a specified margin to the reinsurer, from
which the broker, if any, deducts the brokerage fee. The remain-
ing reinsurance premium is “withheld” by the cedant. The cedant
then pays reinsurance losses out of the funds withheld until they
are exhausted, at which time payments are made directly by the
reinsurer. The reinsurance contract may define a mechanism for
crediting investment income to the funds withheld. The reinsurer
will want a higher profit margin for a funds withheld contract
because of the added risk (the credit worthiness of the cedant)
and the lost investment income.
Additional Cedant Expenses

The cedant incurs various expenses for ceding reinsurance.
These include the cost of negotiation, the cost of a financial
analysis of the reinsurer, accounting, and reporting costs, etc.
If a broker is involved, the brokerage fee covers some of these
services to the cedant. In general, facultative coverage is more
expensive than treaty because of individual policy negotiation,
accounting, and loss cessions.
Reciprocity
In some cases, in order to cede reinsurance, the cedant may be
required to assume some reinsurance from the reinsurer, in this
case usually another primary company. If this reciprocal reinsur-
ance assumption is unprofitable, the loss should be considered
as part of the cost of reinsurance. Reciprocity is not prevalent in
the United States.
Balancing Costs and Benefits

In balancing the costs and benefits of a reinsurance cover
or of a whole reinsurance program, the cedant should consider
more than just the direct costs versus the benefits of the loss cov-
erage and reinsurance functions discussed previously. A major
consideration should be the reinsurer’s financial solidity—Will
the reinsurer be able to quickly pay claims arising from a natu-
ral catastrophe? Will the reinsurer be around to pay late-settled
claims many years from now? Also important is the reinsurer’s
reputation: does the reinsurer pay reasonably presented claims in
a reasonable time? Another consideration may be the reinsurer’s
or broker’s services, including possibly underwriting, marketing,
pricing, loss prevention, claims handling, reserving, actuarial, in-
vestment, and personnel advice and assistance.
Reinsurance Introduction: Final Comments

This introduction is only a brief review of basic reinsurance
concepts and terminology. The interested reader will find more
extensive discussions in the general reinsurance texts listed in
the references.
REINSURANCE PRICING
General Considerations
In general, reinsurance pricing is more uncertain than pri-
mary pricing. Coverage terms can be highly individualized, es-
pecially for treaties. These terms determine the coverage period,
REINSURANCE PRICING 357
definition of “loss,” commission arrangements, premium and loss

payment timing, etc. It is often difficult and sometimes impossi-
ble to get credible loss experience directly pertaining to the cover
being evaluated. Often the claims and exposure data are not as
they first seem. So you must continually ask questions in order
to discover their true nature. Because of these problems of cov-
erage definition and interpretation of loss and exposure statistics,
the degree of risk relative to premium volume is usually much
greater for reinsurance than for primary insurance.
Additional risk arises from the low claim frequency and high
severity nature of many reinsurance coverages, from the lengthy
time delays between the occurrence, reporting, and settlement of
many covered loss events, and also from the leveraged effect of
inflation upon excess claims. In general, the lower the expected
claims frequency, the higher the variance of results relative to
expectation, and thus the higher the risk level. In addition, IBNR
claims emergence and case reserve development are severe prob-
lems for casualty excess business. Claims development beyond
ten years can be large, highly variant, and extremely difficult
to evaluate, as we shall discuss in the Loss Reserving section.
Because of this long tail and extreme variability of loss pay-
ments, the matching of assets with liabilities is more difficult.
Future predictability is also decreased by the greater uncertainty
about claims severity inflation above excess cover attachment
points. All these elements create a situation where the variance
(and higher moments) of the loss process and its estimation are
much more important relative to the expected value than is usu-
ally the case for primary coverage. For some reinsurance covers,
the higher moments (or at least the underwriter/actuary’s be-
liefs regarding uncertainty and fluctuation potential) determine
the technical rate.
Reinsurance Pricing Methods

There are many ways to price reinsurance covers. For any
given situation, there is no single right way. In this section, we
will discuss a few actuarially sound methods. In general, the

exposition of pricing methods will begin simply and become
more complex as the situation demands and as we ask more
questions. In many real situations, you might want to get a quick
first evaluation via the simplest methods. Indeed, if you judge
the situation to be either fairly predictable by these methods, or
if you judge the risk to be small, you may decide to stop there. If
not, you may want to pursue your analysis and pricing along the
lines presented here. As in most actuarial work, you should try
as many reasonable methods as time permits (and also reconcile
the answers, if possible).
In this spirit, please note that the simple flat rate pricing for-
mula and the use of the Pareto and Gamma distributions in this
chapter are for illustrative purposes. The pricing formula a rein-
surance actuary would use depends upon the reinsurer’s pric-
ing philosophy, information availability, and complexity of the
coverage. The probability models should be selected to describe
the actual situation as best as possible given all the real statisti-
cal and analytical cost constraints preventing you from obtaining
more information. Klugman, Panjer, and Willmot [12], Hogg and
Klugman [11], and Patrik [18] all discuss model selection and
parameter estimation.
A Flat Rate Reinsurance Pricing Formula

As we get into the formulas, there will be a lot of notation. For
clarity, we will preface variables with PC for “primary company”
and R for “reinsurer” or “reinsurance.” PV will be used in the
traditional sense to mean “present value.”
A discussion of the pricing formula to be used in this chapter
will illustrate certain differences from primary pricing. We will
often use the word “technical” to distinguish the actuarially cal-
culated premium, rate, etc. from the actual final premium, rate,
etc., agreed to by the cedant and reinsurer. Formula 7.2 calculates
the technical reinsurance premium in terms of reinsurance loss
cost, external expenses, internal expenses, and target economic
return in the simple case where there are no later commission

or premium adjustments based upon the actual loss experience.
You can see that 7.2 is an “actuarial” formula, based explicitly
upon costs.1
Formula 7.2: A Flat Rate Reinsurance Pricing Formula
PVRELC
RP =
(1 ! RCR ! RBF) " (1 ! RIXL) " (1 ! RTER)
where
RP = reinsurance premium
PVRELC = PV of RELC
= RDF " RELC
RELC = reinsurer’s estimate of the reinsurance expected
loss cost, E[RL]
RL = reinsurance loss
E[RL] = reinsurance aggregate loss expectation
RDF = reinsurance loss payment discount factor
RCR = reinsurance ceding commission rate
(as a percent of RP)
RBF = reinsurance brokerage fee
(as a percent of RP)
RIXL = reinsurer’s internal expense loading
(as a percent of RP net of RCR and RBF)
RTER = reinsurer’s target economic return
(as a percent of reinsurance pure premium,
RP net of RCR, RBF and RIXL)
1 Formulas traditionally used by reinsurance underwriters have more often been of the
form: undiscounted loss estimate divided by a judgmental loading factor such as 0.85.
Of course the problem with this type of formula is that all the information about the
expenses, discounting and profit loading are buried in one impenetrable number.
Let’s break down this somewhat complicated-looking for-

mula. First, as an actuarial technical pricing formula, we build
up the premium starting with the loss cost. The reinsurance pure
premium (RPP) can be written as follows.
Formula 7.3: Reinsurance Pure Premium
PVRELC
RPP =
1 ! RTER
This is the real “risk” premium for the risk transfer. The rein-
surer’s target economic return, RTER, is the reinsurer’s charge
for profit and risk. It is properly related to PVRELC for the
contract. By writing the formula this way, the reinsurer’s ex-
pected profit (present value) is RTER " RPP. In this flat rated
case, the reinsurer’s expected profit is a simple linear func-
tion of the present value of the reinsurance expected loss cost,
RTER=(1 ! RTER) " PVRELC. A discussion of how to select
RTER’s for various contracts or lines of business is well beyond
the scope of this chapter. The modern actuarial and economic
literature derives these profit margins from an overall corpo-
rate target return on equity (RoE) and the relative risk level of
each particular contract. There is extensive actuarial literature on
this topic; see especially Bühlmann [9], Daykin, Pentikäinen and
Pesonen [8], and, generally, The ASTIN Bulletin, The Proceedings
of the Casualty Actuarial Society and any publications connected
with the CAS Casualty Actuaries in Reinsurance.
Next, the reinsurer’s internal expenses are loaded onto the
reinsurance pure premium by dividing it by 1 ! RIXL. Thinking
of this from the opposite direction (top down), it may also be
thought of as a loading on reinsurance premium less external ex-
penses. It is convenient to think of the loading for the reinsurer’s
internal expenses this latter way for at least three reasons:
1. This is the reinsurer’s actual cash income from the cedant
(unless there are funds withheld).
2. It is relatively easy to account for external expenses by
reinsurance line of business. Within a line, the reinsurer’s
underwriting and claims handling effort, and thus inter-

nal expenses, should be similar for each contract, varying
only by claims and “risk” volume.
3. Internal expenses by contract should be independent (or
almost independent) of commissions or brokerage ex-
penses. Thus the loading should be independent of these
expenses.
Together, RTER and RIXL determine the reinsurer’s desired
pricing margin for internal expenses and profit (economic re-
turn).
Finally, the reinsurance ceding commission rate and broker-
age fee are specified in each particular contract; they are almost
always stated as percentages of the total reinsurance premium
RP. There is often no ceding commission on excess coverage;
this is very different from primary coverage where commissions
almost always exist. Of course, the existence of a reinsurance
brokerage fee also depends upon the existence of a reinsurance
intermediary for the contract.
An example will help clarify this.
Example 7.4:
PVRELC = $100,000 (calculated by actuarial analysis

and formulas)
RTER = 20% (The reinsurer believes this is appropriate
to compensate for the uncertainty and risk level
of the coverage.)
RIXL = 10% (The reinsurer’s allocation for this type
of business.)
RCR = 25% (specified in the contract)
RBF = 5% (specified in the contract)
Then
$100,000
RPP = = $125,000
0:8
$125,000
RP =
(1 ! :10) " (1 ! :25 ! :05)
= $198,413
Please note that the reinsurance premium less external ex-
penses is
RPP $125,000
=
1 ! RIXL 0:9
= $138, 889
= 0:7 " $198,413
= (1 ! RCR ! RBF) " RP
Also, the reinsurer’s desired margin for internal expenses and
profit is
$138,889 ! $100,000 = $38,889
= $13,889 + $25,000:
Very often the reinsurance premium is not a fixed dollar
amount, but is calculated as a rate times a rating basis, quite
often PCP, the primary company (subject) premium. In our ex-
ample, if PCP was expected to be $5,000,000, the reinsurance
rate would most likely be rounded to 0.04 or 4%. Then the
expected reinsurance premium would be $200,000 and the ex-
pected reinsurance premium less external expenses would be
0:7 " $200,000 = $140,000, and the expected reinsurance mar-
gin would be $140,000 ! $100,000 = $40,000, greater than the
desired margin. So, if the reinsurer’s internal expenses were still
$13,889, then the reinsurer’s expected economic return (profit)
would be $40,000 ! $13,889 = $26,111, greater than the target
of $25,000.
Later we will see cases where the commission or premium is

dependent upon the actual loss experience on the contract. This
leads to a more complicated formula, where in order to obtain
a proper actuarial technical premium or rate, you must consider
the reinsurer’s profit as a more complicated function of the loss
experience. But let us start simply.
Thirteen-step Program to Reinsurance Pricing Happiness

In general, when pricing reinsurance, it is desirable to perform
both an exposure rating and an experience rating. An exposure
rate is akin to a primary manual rate, using general rating fac-
tors independent of the cedant’s particular claims experience.
An experience rate is akin to a primary loss rate, completely
dependent upon the cedant’s particular claims experience. The
final technical rate will be a weighing together of both of these
rates.
The steps in the rating procedure may be abstracted as follows.
1. Gather and reconcile primary exposure, expense and

rate information segregated by major rating class
groups.
2. Calculate an exposure expected loss cost, PVRELC, and,
if desirable, a loss cost rate, PVRELC/PCP.
3. Gather and reconcile primary claims data segregated by
major rating class groups.
4. Filter the major catastrophe claims out of the claims
data.
5. Trend the claims data to rating period.
6. Develop the claims data to settlement values.
7. Estimate the catastrophe loss potential.
8. Adjust the historical exposures to the rating period.
9. Estimate an experience expected loss cost, PVRELC,

and, if desirable, a loss cost rate, PVRELC/PCP.
10. Estimate a “credibility” loss cost or loss cost rate from
the exposure and experience loss costs or loss cost rates.
11. Estimate the probability distribution of the aggregate
reinsurance loss, if desirable, and perhaps other distri-
butions, such as for claims payment timing.
12. Specify values for RCR, RIXL, and RTER.
13. Negotiate, reconcile opinions and estimates, alter terms,
and finalize.
Steps 1, 2, and 12 may be considered to be exposure rating,
steps 3–9 and 12 to be experience rating, and steps 10–13 to
be rate finalization. Step 11 is usually performed only for more
complex contracts. We will try to use this same sequence of
steps whenever possible. But sometimes the order of the steps
will differ depending upon the particular situation. Let us start
with the simplest case.
Facultative Certificates
Since a facultative certificate covers a share of a single in-
surance policy or set of policies covering a single insured, the
individual insured can be underwritten and priced. The exposure
of the individual insured can be evaluated and manual rates and
rating factors can be used to calculate an exposure rate. How-
ever, since most facultative certificates are written on larger or
more hazardous exposures, manual rates and rating factors may
not exist or must often be modified. Thus, the analysis of indi-
vidual exposure and loss experience, together with a great deal
of underwriting judgment, is important.
In contemplating any form of facultative coverage, the under-
writer first evaluates the exposure to decide if the risk is accept-
able, and may then evaluate the rate used by the cedant to esti-
mate its degree of adequacy. The underwriter also determines if
the ceding commission fairly covers the cedant’s expenses, but

does not put the cedant into a situation significantly more ad-
vantageous than that of the reinsurer. That is, except in unusual
circumstances, the cedant should not make a significant profit
in those circumstances where the reinsurer is losing money, and
vice versa.
The Actuary’s Role

Historically, actuaries have seldom been directly involved in
facultative certificate pricing. But they can be useful in the fol-
lowing ways by:
1. Being sure that the facultative underwriters are provided

with, and know how to use, the most current and accu-
rate manual rates and rating factors, e.g., increased limits
factors, loss development factors, trend factors, actuar-
ial opinions on rate adequacy by exposure type and by
territory (state), etc.
2. Working with the underwriters to design and maintain
good pricing methodologies, almost always in the form
of interactive computer programs.
3. Working with the underwriters to design and maintain
good portfolio monitoring systems for meaningful cate-
gories of their business, for relative price level and for
the monitoring of claims, underwriting, and bottomline
profit experience.
4. Working with the underwriters to evaluate and determine
which lines of business and which exposure layers to
concentrate upon as market conditions change.
Property Certificates
The evaluation and pricing of property certificate coverage on
a proportional share basis usually needs little further actuarial
assistance. However, the actuary should be involved in the eval-
uation of the accumulation of catastrophe exposure, and also in

the interpretation of long-term results from a corporate portfolio
perspective.
The evaluation and pricing of property certificate coverage
on an excess basis is more difficult. Many tables used by un-
derwriters calculate the excess rate as a factor times the primary
rate.2 Instead of relating the excess rate directly to the primary
rate, a more actuarially sound method for determining property
per-risk excess rating factors expresses the expected excess loss
cost for coverage above an attachment point as a percentage of
the total expected loss cost. This allows us to use Formula 7.2
to treat expected loss cost, expenses, and profit separately. The
curves underlying these factors depend upon the class of business
(its claim severity potential) as determined by the most impor-
tant rating variables: amount insured, MPL (maximum possible
loss), PML (probable maximum loss), construction, occupancy,
and protection.
The MPL, sometimes called the “amount subject,” is a very
conservative estimate by the individual underwriter of the max-
imum loss possible on the policy. For example, it includes the
maximum full value of contiguous buildings together with con-
tents, and also reflects maximum time element (e.g., business in-
terruption) coverage. The PML is a less conservative estimate of
the largest loss, assuming for example, that the sprinkler system
works, that the contents are normal, etc. The difference between
MPL and PML is illustrated by considering an office building:
2 Some underwriters use so-called Lloyd’s Scales. Underwriters bring these tables with
them from job to job; the parentage of the Lloyd’s Scales floating around the industry
seems to be highly questionable. So, be careful.
Some underwriters also use tables of factors from a 1960 PCAS paper by Ruth Salz-
mann. But these factors were developed for homeowners business. So even if they were
adjusted for inflation over the last 40 years, they are of questionable use for the typical
commercial property facultative exposure. A paper by Stephen Ludwig [13] uses com-
mercial risk experience to estimate loss curves. ISO and various reinsurance companies
have been doing research and developing property loss curves. But so far there are no
published, actuarially sound claims severity curves or tables of factors for rating property
per-risk excess coverage.
the MPL is the total value; the definition of PML varies from
underwriter to underwriter, but is usually thought to be three
to five floors. The MPL and PML affect the shape of the loss
cost curve because you expect, for example, very different loss
severity distributions for an insured with a $100,000 MPL and
PML versus an insured with a $10,000,000 MPL and $5,000,000
PML.
This is illustrated by the accompanying Graph 7.5. An actu-
ary might think of the MPL as being essentially the 100th per-
centile of the probabilistic loss distribution, and the PML as be-
ing somewhere around the 95th to 99th percentiles. Note that the
$10,000,000 MPL property has a smaller probability of a total
loss. In fact, at every loss level, its graph lies above the graph for
the $100,000 MPL property, thus having a smaller probability of
exceeding each percentage loss level.
Appropriate RTER’s and RIXL’s could be incorporated into
the table or could be recommended as additional loading fac-
tors.
An appropriate pricing formula for an excess cover could use
Formula 7.2 with (dropping the PV for short-tailed property cov-
erage) RELC calculated as follows.
Formula 7.6: Reinsurance Expected Loss Cost

RELC = ELCF " PCP " PCPLR " RCF
where
ELCF = excess loss cost factor
(from the table; as a percent of total loss cost)
PCP = primary company (subject) premium
PCPLR = primary company permissible loss ratio
(including any loss adjustment expenses covered
as part of loss)
GRAPH 7.5
Examples of Property Claim Severity
Cumulative Distribution Functions
RCF = rate correction factor

(reinsurer’s adjustment for the estimated
(in)adequacy of the primary rate)
Again, the reinsurance premium can be translated into a rein-
surance rate by dividing it by the primary company premium
PCP. An example will help clarify this.
Example 7.7:
PCP = $100,000
PCPLR = 65%
RCF = 1:05 (estimated 5% inadequacy)
RCR = 30%
RBF = 0% (no broker)
RIXL = 15%
MPL = PML = $10,000,000:
Attpt = Attachment Point = $1,000,000:
RLim = Reinsurance Limit = $4,000,000
RTER = 10%
Thus the reinsurer believes the total expected loss cost is as
follows.
Formula 7.8: Total Expected Loss Cost
PCP " PCPLR " RCF = $100,000 " (:65) " (1:05)
= $68,250
Now assume that we believe that the claim severity, includ-
ing loss adjustment expense, for this class of business and this
MPL, is given by a censored (at MPL) Pareto distribution of the
following form.
Formula 7.9: Censored Pareto Model
!b
" for x<1
1 ! F(x) = Prob[X > x] = (b + x)q
#
0 for x#1
where the claim size X is expressed as a percent of MPL.
(Properties of the Pareto distribution are outlined in Appen-
dix A.)
Suppose that the parameters are given by b = 0:1 and q = 2.

We can now verify the following facts.
Formula 7.10: Expected Claim Severity (as a fraction of MPL)

$ % & ' (q!1 )
b b
E[X; 1] = " 1! (Appendix A)
q!1 b+1
$ % & ' (1 )
0:1 0:1
= " 1!
2!1 1:1
= $0:1% " $1 ! 0:91%
= 0:0909
Thus, if a loss occurs, our estimate of its average size is
0:0909 " $10,000,000 = $909,000.
We can also calculate the expected claim count.
Formula 7.11: Expected Claim Count (ground-up)

Total expected loss cost
Expected claim count =
Expected loss severity
$68,250
= = 0:075
$909,000
Formula 7.12: Expected Claim Severity in the Layer

[$0, $1,000,000]
$ % & ' (q!1 )
b b
E[X; 0:1] = " 1!
q!1 b + 0:1
$ % & ' (1 )
0:1 0:1
= " 1!
2!1 0:1 + 0:1
= $0:1% " $1 ! 0:5%
= 0:05
Thus, if a loss occurs, its average size in the layer
[$0, $1,000,000] is 0:05 " $10,000,000 = $500,000:

[$0, $5,000,000]
$ % & ' (q!1 )
b b
E[X; 0:5] = " 1!
q!1 b + 0:5
$ % & ' (1 )
0:1 0:1
= " 1!
2!1 0:1 + 0:5
= $0:1% " $1 ! 0:167%
= 0:0833
Thus, if a loss occurs, its average size in the layer
[$0, $5,000,000] is 0:0833 " $10,000,000 = $833,000: There-
fore, the ELCF can be calculated as follows.
Formula 7.14: ELCF (Excess Loss Cost Factor)3
E[X; 0:5] ! E[X; 0:1]
ELCF =
E[X; 1]
:0833 ! 0:5
=
0:0909
= 0:367
We can now calculate the Reinsurance Expected Loss Cost
(RELC).
Formula 7.15: Reinsurance Expected Loss Cost
RELC = est. E[RL] = ELCF " $Total expected loss cost%
= 0:367 " $68,250
= $25,028
3 The factors in a Lloyd’s-type table, mentioned in a previous footnote, would technically

be of the following form:
Expected loss cost fraction at limit c% of total amount insured
E[X; c]
= for 0 & c & 1
E[X; 1]
Thus the ELCF is simply a difference of two of these factors: at the attachment point
and at the attachment point plus the limit.
Then we can calculate the technical reinsurance pure premium

as
RELC
RPP =
1 ! RTER
$25,028
=
0:9
= $27,808
and the technical reinsurance premium as
RPP
RP =
(1 ! RIXL) " (1 ! RCR ! RBF)
$27,808
=
(0:85) " (0:7)
= $46,737:
To squeeze out more information about the reinsurance cov-

erage in order to measure the risk transfer, as we will discuss
later, we can also estimate the reinsurance expected claim count
and severity.
Formula 7.16: Expected Reinsurance Claim Count
Expected excess claim count

= Expected claim count (ground-up) " Excess probability
' (q
b
= 0:075 "
b + 0:1
' (2
0:1
= 0:075 "
0:1 + 0:1
= 0:075 " 0:25
= 0:0188
We can then calculate the excess expected claim severity.


[$1,000,000, $5,000,000]
Excess expected claim severity

RELC
=
expected reinsurance claim count
$25,028
=
0:0188
= $1,333,333:
This is the average size (in the layer) of a claim entering the
layer. Note that it is one third of the reinsurance limit. Average
intra-layer claim sizes for excess layers are typically (but not
always) between one quarter to one half of the layer limit.
Further Property Certificate Pricing Considerations

Quite often the pricing situation is much more complicated,
with multiple locations and coverages. The underwriter/pricer
generally determines a price for each location and each coverage,
and then adds them to obtain the total premium.
Instead of working directly with an estimated underlying
loss severity distribution like this Pareto, the ELCF in For-
mula 7.14 might be obtained from a Lloyd’s-type table. Bet-
ter yet, a complete pricing procedure such as this can be pro-
grammed into an interactive computer package for the under-
writers. The package would contain all the appropriate rates
and rating factors or underlying loss severity models and pa-
rameters to be called upon by the user. It would ask most
of the relevant questions of the user and would document the
decision trail for each submission seriously contemplated by
the underwriter. If desirable, the reinsurer’s values for RIXL’s
and RTER’s could be built into the system or as part of the
ELCF’s.
For facultative certificate property coverage as with any rein-
surance business segment, the pricing cycle is very severe. This
is mainly due to naive capital flowing into the market because of

easy access, but also due to the short-term nature of most peo-
ples’ memories. Thus it is very important to monitor the results
closely. Renewal pricing and rate competition in the marketplace
should be watched monthly; perhaps summaries derived from the
aforementioned pricing system would be appropriate. Quarterly
updates of underwriting results by accident year in appropriate
business segment detail are very important.
Casualty Certificates
The evaluation and pricing of facultative certificate casualty
coverage is even trickier than property coverage, due mainly to
the additional uncertainty arising from delayed claims report-
ing and settlement. Because of this increased uncertainty, the
actuary’s role can be more important in the pricing and in the
monitoring and interpretation of results.
As with property excess, a cover may be exposure rated via
manual rates and increased limits factors, together with exposure
evaluation and underwriting judgement. The same Formula 7.6
may be used to determine RELC, except that the ELCF will be
based upon increased limit loss cost tables, based upon claim
severity curves, and the RCF may be determined both by facts
and by judgments regarding the cedant’s basic limit rate level
and increased limit factors.
Since most companies use Insurance Services Office (ISO)
increased limit factors for third party liability pricing (especially
for commercial lines), it is very important that the actuaries very
closely monitor ISO factors and understand their meaning. Like-
wise, it is important to monitor and understand information from
the National Council on Compensation Insurance (NCCI) regard-
ing workers compensation claim severity. However, you should
not use the published excess loss factors (ELF’s) for pricing ex-
cess loss coverage, since they underestimate per-occurrence ex-
cess loss potential.
Allocated Loss Adjustment Expense (ALAE)

ALAE per claim is usually covered on an excess basis either:
' In proportion to the indemnity loss share of the excess cover
vis-a-vis the total, or
' by adding the ALAE to the indemnity loss before applying the
attachment point and limit.
For example, assume the layer $500,000 excess of $500,000 is
reinsured. The following claims induce the indicated reinsurance
reimbursements.
TABLE 7.18
Reinsurance Payments by ALAE Share
(1) (2) (3) (4)

Indemnity Reinsurer’s payment Reinsurer’s payment
Payment ALAE if ALAE pro rata if ALAE added
$ 500,000 $ 50,000 $0 $ 50,000

750,000 75,000 275,000 325,000
1,000,000 100,000 550,000 500,000
2,000,000 150,000 537,500 500,000
Increased limits factors published by ISO have no provision

for ALAE outside of the basic limit. Thus ELCF’s based upon
these increased limit factors must be adjusted to cover the rein-
surer’s share of ALAE.
Pricing Methods
Since policies subject to facultative coverage are larger than
usual, experience rating often comes into play. One method is to
first experience rate a lower layer with more credible experience.
Then the experience-based loss cost on the lower layer may be
used together with the reinsurer’s ELCF table to extrapolate up
to the intended layer of coverage.
For a buffer layer of coverage where the likelihood of loss
penetration is significant, it might also be possible to obtain a
loss cost estimate directly from a careful analysis of the large

loss experience of the insured.
An example should clarify this.
Example 7.19:
' Policy period priced = 2001

' Exposure = General liability premises/operations
' Policy limit = $2,000,000, no aggregate
' PCP = $550,000 (estimated for 2001)
' BLP (basic limit premium @ $100,000 limit) = $200,000 (es-
timated for 2001)
' PCPLR = 70% (excluding unallocated loss adjustment expense
from the loss cost)
' Attpt (attachment point) = $500,000
' RLim (reinsurance limit) = $1,500,000
' ALAE is covered pro rata.
' RCR (reinsurance commission) = 25%
' RBF (reinsurance brokerage fee) = 5%
' RIXL (reinsurer’s internal expense loading) = 15%
' RTER (reinsurer’s target economic return) = 20%
' Have exposure and loss experience for policy years 1995
through 1999, consisting of exposures, basic and total lim-
its premiums, current evaluation of basic limits losses and a
detailed history for each known claim larger than $100,000
Suppose that the cedant’s basic limit premium was determined
from a standard experience and schedule rating plan that we
believe to be adequate. Also suppose that the cedant uses the
appropriate ISO increased limit factors, which we believe also
to be adequate, and which include the ISO risk loading but no

ALAE provision for the layer. Suppose the ISO increased limit
factors for this exposure are as follows.
TABLE 7.20
(Fictitious) ISO Increased Limit Factors
(1) (2) (3)

Policy limit Published ILF ILF without risk load
$ 100,000 1.00 1.0000

500,000 2.10 1.9077
2,000,000 2.75 2.4264
Suppose that the cedant is offering a manual difference excess

premium of $130,000 calculated as follows.
Formula 7.21: Manual Difference Excess Premium
Manual difference excess premium = $200,000 " (2:75 ! 2:10)

= $130,000
This is the simplest technical price determination possible.
Some reinsurers, after multiplying by a judgment factor to adjust
for rate (in)adequacy for this insured or for this type of insured,
stop here. However, because there is too much information hid-
den in this calculation, let us continue onward to calculate an
actuarial, technical price based upon the estimated loss cost and
the other rating parameters.
Suppose that, based upon a study of the relationship of ALAE
to claim size for this type of exposure, we believe that an ap-
propriate loading for pro rata ALAE is 10% of indemnity loss
cost for this layer. Also suppose that we have an estimated ex-
pected loss payment pattern for this type of exposure and this
layer, and suppose that the corresponding discount factor, using
current U.S. Treasury rates (risk-free) timed to the loss payment
pattern and reflecting the implications of the current tax law, is

.80. Then if we believe the ISO increased limit factors to be ad-
equate for this exposure, the present value reinsurance expected
loss cost could be calculated as follows.
Formula 7.22: PVRELC
PVRELC = RDF " RELC

= RDF " ELCF " BLP " PCPLR " RCF
= RDF " ELCF " $basic limit expected loss cost%
= (0:8) " $(2:4264 ! 1:9077) " (1:10)%
" $$200,000 " (0:7) " (1:00)%
= (0:8) " $(0:5187) " (1:10)% " $$140,000%
= 0:8 " 0:5706 " $140,000
= $63, 904
Then the reinsurance premium can be calculated via Formula
7.2.
Formula 7.23: Reinsurance Premium (RP)
PVRELC
RP =
(1 ! RCR ! RBF) " (1 ! RIXL) " (1 ! RTER)
$63,904
=
(1 ! 0:25 ! 0:05) " (1 ! 0:15) " (1 ! 0:2)
$63,904
=
(0:7) " (0:85) " (0:8)
$63,904
=
0:476
= $134, 252
Please note that the assumption that ISO increased limit fac-
tors appropriately describe the claim severity potential for this
insured is a very crucial assumption. Since facultatively rein-

sured exposures are often more hazardous, the actuary or under-
writer may believe that often the claim severity distribution is
more dangerous. However, the actuary or underwriter designing
a facultative certificate pricing procedure may believe that the in-
creased severity hazard is already reflected in the first $100,000
basic limit price for these insureds. Otherwise, you may wish to
adjust ISO claim severity curves accordingly, or allow the cer-
tificate pricer to do so on a case-by-case basis.
In this case, the offered $130,000 premium looks reasonably
close to our technical premium. So the pricing endeavor usually
stops here. But what about the large loss experience? Suppose
that for accident years 1995–1999, there are some claims known
as of June 30, 2000 whose indemnity values are greater than
$100,000. Can any of this insured’s large loss information be
used to help price the cover, or at least to help form an opinion
as to the adequacy of the exposure rate premium?
A recommended rating procedure in this case is to experi-
ence rate the layer $400,000 excess of $100,000, and then use
the non-risk-loaded increased limit factors to extrapolate upward
to a reinsurance loss cost for the layer $1,500,000 excess of
$500,000. We will describe a common reinsurance experience
rating methodology later in the section on pricing excess treaties.
As with property excess, it is clear that the exposure rating
methods can be programmed into an interactive computer pack-
age for underwriters. Also, as with property coverage, it is very
important to monitor relative rate level and results in appropriate
business segment detail. The actuarial evaluations and opinions
regarding future case reserve development and IBNR emergence
should be very important to the underwriters.
Facultative Automatic Programs

These large multi-insured programs are very similar to
treaties. One difference, however, is that the reinsurance premium
for a facultative automatic excess cover is usually computed on

a policy-by-policy basis using agreed-upon excess rates, instead
of as a rate times total subject premium, as is usually the case
for treaties. Thus the reinsurance premium is more responsive
to the individual exposures ceded to the reinsurer. Nevertheless,
the risk of anti-selection against the reinsurer on a nonobligatory
contract should be evaluated by the underwriter.
The pricing of these agreements is the same or similar to the
pricing of excess treaties, discussed below.
Reinsurance Treaties in General

In this discussion the word “treaty” is used interchangeably
for both treaties and automatic facultative agreements.
Since a treaty covers a share of a portfolio of insurance poli-
cies, insureds are rarely individually underwritten and priced by
the reinsurer; many of the policies to be covered may not be writ-
ten by the cedant until well into the reinsurance coverage period.
Instead, the reinsurance underwriter/pricer considers the whole
portfolio of potentially subject policies. To do this, the reinsurer
evaluates first the management of the potential cedant. What is
their management philosophy and ability? Are they honest, fair-
dealing? Do they know what they are doing? Is the company
financially solid? What are their business plans? Why do they
want reinsurance? Why do they need reinsurance?
Once the reinsurer is satisfied that this is a company and these
are people we would like to deal with on a long-term basis, we
can then evaluate their underwriting, primary pricing, market-
ing, and claims handling ability. Since individual insureds are
not usually underwritten by the reinsurer, we must be generally
satisfied with the cedant’s underwriting and pricing expertise for
the exposure we may assume. For any treaty, we must under-
stand the cedant’s insurance exposures, rate level, and policy
limits sold. Many reinsurers will send a team of marketing and
underwriting people to perform a pre-quote audit, and will also
send claims people to review the company’s claims handling and

reserving practices.
The reinsurer (or the lead reinsurer on a multi-reinsurer bro-
kered treaty) usually reviews the structure of the cedant’s rein-
surance program to understand how all the reinsurance con-
tracts, facultative and treaty, fit together to provide benefits to
the cedant. Lastly, the reinsurer evaluates the particular reinsur-
ance treaties and suggested rates if offered, or creates a program
and rates to offer to the cedant company.
Actuaries can provide extremely useful, and often necessary,
technical support for treaty business. They can perform the four
functions listed at the beginning of the section on facultative
certificate pricing. They can also get involved in the technical
evaluation and pricing of individual large and/or difficult treaties
through many or all of the rating steps. Experience rating is much
more important for treaties than for facultative certificates. Con-
sequently, the actuarial tools of data analysis and loss modeling
can be critical to a reinsurer’s ability to write difficult exposures,
especially casualty exposures where long-tail loss development
is a significant factor.
Property Proportional Treaties

Proportional treaties are usually priced by evaluating the
amount of the commission to be paid ultimately to the cedant.
The ultimate commission is comprised of a provisional commis-
sion paid at the time of the cession, plus any later adjustments
specified by the terms of the treaty.
Property Quota-Share Treaties

A traditional quota-share treaty covers a share of the cedant’s
net retention after facultative covers. To evaluate the loss ex-
posure, we follow the rating steps outlined before. To facilitate
the discussion, we will consider an example using our favorite
primary insurer, da Ponte Insurance Company.
Example 7.24:
' Rating period = 2001

' 25% quota share on various property lines
' Risks attaching coverage (primary policy year; reinsurers gen-
erally call this “underwriting year”)
' Estimated PCP (written premium) = $10,000,000
' Estimated unearned premium reserve@12=31=01=$3,000,000
' Per occurrence limit of $7,500,000 for reinsurance losses
' Proposed RCR = 35%
' RBF = 0% (no broker)
Step 1: Gather and reconcile primary exposure, expense and rate

information segregated by major rating class groups.
The grouping may be by Annual Statement line of business,
or could be a finer decomposition if the proposed reinsurance
coverage is more restricted or if there are any important expo-
sures you may want to evaluate separately. What is meant by
the word “exposure” for reinsurance purposes is usually primary
subject premiums. The reconciliation should be to publish finan-
cial records as much as possible. If there is significant catastrophe
potential, we would want the exposure by zip code in order to
perform the seventh rating step.
Let us suppose that the proposed treaty is to cover the property
exposures in Annual Statement lines 1–5 and 12, net of faculta-
tive reinsurance; the liability parts of lines 3–5 are excluded. In
this example, the reinsurer would ask for gross written premium
by line by year for 1995 through 6/30/00, together with estimates
for 2000 and 2001. The expense information could be from the
cedant’s Insurance Expense Exhibit. The rate information would
be contained in the cedant’s underwriting line guide. We also
want information on average rate deviations.
Step 2: Calculate an exposure expected loss cost, PVRELC, and,

For proportional coverage, an exposure loss cost rating is sim-
ply an evaluation of the adequacy of the cedant’s rates for the ex-
posures to be covered, leading to an estimate of the expected loss
ratio. An underwriting review can compare the cedant’s rates to
those of other primary insurers or to the reinsurer’s own database
of adequate primary rates.
Many people in the reinsurance business would say that you
cannot calculate an exposure rate for proportional coverage, or
that you cannot rate the coverage at all; you can only evaluate
the ceding commission. The ceding commission should fairly
reimburse the cedant’s expenses, but should not put the cedant
into a position significantly more advantageous than that of the
reinsurer. Except in unusual circumstances, the cedant should not
make a significant profit while the reinsurer is losing money, and
vice versa. The point here is to evaluate the (in)adequacy of the
cedant’s rates in order to evaluate whether or not the proposed
reinsurance commission will work for the reinsurer and for the
cedant.
Let us suppose that the review indicates that overall, the ex-
pected loss ratio for policy year 2001 is 65% including all catas-
trophe losses. Suppose we estimate that a per occurrence catas-
trophe limit will only reduce this by a few percent, but less than
the amount of the reinsurer’s expense and profit loadings, RIXL
and RTER. Thus, it looks as if the reinsurer cannot afford a 35%
ceding commission. We will deal with the ceding commission
issue further in steps 10–12.
Step 3: Gather and reconcile primary claims data segregated by

We want the claims data segregated as the exposure data. We
usually want ALAE separately from indemnity losses. For propor-
tional coverage, the data are usually historical aggregate claims
data for the past five to ten policy years, plus individual large
claims. We also want some history of claims values (an histori-
cal policy year/development year triangle) for step five. The data
should be adjusted so that they are approximately on the same
basis as our coverage with respect to any other inuring reinsur-
ance, that is, reinsurance that applies to the cedant’s loss before
our coverage.
Suppose we have net aggregate loss development triangles by
line for policy years 1995–2000 at evaluation dates 12/31/95,
12/31/96, : : : ,12/31/99, 6/30/00, plus concurrent evaluations by
line of all numbered catastrophes occurring during this time pe-
riod. These are the catastrophes designated by Property Claims
Service (PCS). Also suppose that there haven’t been significant
changes in da Ponte’s insurance portfolio or in their reinsurance
program during this time. So the data are consistent with the
expected 2001 loss exposure.
Step 4: Filter the major catastrophic claims out of the claims data.
This is straightforward. Subtract the numbered catastrophe
values by line at each evaluation from the loss development tri-
angles.
Step 5: Trend the claims data to the rating period.

We want to be sure that the historical claims data are adjusted
in a manner that makes them reasonably relevant to the rating
period. The trending should be for inflation and for other changes
in exposure (such as higher policy limits) that might affect the
loss potential. For proportional coverage, we may skip this step,
and simply look for any apparent trend in the historical loss ratios
in step nine.
Step 6: Develop the claims data to settlement values.

Claims development is usually not much of a problem for
filtered primary property claims. If we have historical policy
year/development year triangles, we can use standard methods

to estimate loss development factors and apply these to the fil-
tered data. If we don’t have historical triangles, we may be able
to use Annual Statement Schedule P accident year data, if they
are reasonably reflective of the proposed coverage and if we can
filter out major catastrophes, to estimate policy year loss develop-
ment factors. We also want to compare the development patterns
estimated from the cedant’s data to what we expect based upon
our own historical data for comparable coverages to check for
reasonableness.
If the reinsurer believes that the development patterns should
be reasonably similar for the various covered lines, we usually
estimate the total development from the combined data instead
of by line.
Step 7: Estimate the catastrophic loss potential.

This deserves a whole section by itself. The problem is that
the historical data, even developed and trended, may not indicate
the true potential for catastrophic losses. Reinsurers who have re-
lied entirely upon five to ten years of historical claims to estimate
catastrophe potential have often blundered into coverages where
the catastrophic events occurring during their coverage period
have more than wiped out the “normal” non-catastrophic prof-
its. For example, da Ponte’s 1995–1999 experience period does
not include such major catastrophes as Hurricane Andrew (1992)
and the Northridge earthquake (1994). It also does not include
such possibilities as a New Madrid fault earthquake, a Long Is-
land hurricane, etc. If da Ponte’s exposures include any poten-
tial catastrophe areas, then we must estimate the corresponding
catastrophe potential.
We will discuss various methods for this evaluation in the
later section on catastrophe covers. If we collected gross written
premiums by zip code, we can use these data to parameterize
one of the available catastrophe computer models.
Let us assume that our evaluation of this portfolio indicates

an expected catastrophic loss ratio to PCP of 12% with respect
to a $7,500,000 per occurrence limit on reinsurance losses.
Step 8: Adjust the historical exposures to the rating period.

We want to be sure that the historical exposure (premium) data
are adjusted in a manner that makes them reasonably relevant
to the rating period. The trending should be for primary rate
and underwriting changes and for other changes in exposure that
might affect the loss potential. For proportional coverage, we
only adjust the premiums for significant rate level changes, so
that the historical loss ratios in step nine are all consistent with
2001 rate level. The rate adjustments are accomplished using the
standard methods.
Step 9: Estimate an experience expected loss cost, PVRELC, and,

Suppose that our data and estimates so far are as displayed in
Table 7.25.
TABLE 7.25
Da Ponte Insurance Company Property Business
(in $1,000’s)
(1) (2) (3) (4) (5) (6) (7) (8)

Loss Devel’d Loss
Policy Onlevel Subject Cat. Filtered Devel. Loss Ratio
Year PCP Loss Loss Loss Factor (5) " (6) (7) ( (2)
1995 $ 7,000 $ 3,472 $ 512 $ 2,960 1.00 $ 2,960 42%
1996 7,500 4,116 403 3,713 1.00 3,713 50%
1997 8,000 4,772 188 4,584 1.01 4,630 58%
1998 8,500 4,855 1,286 3,569 1.05 3,747 44%
1999 9,000 4,144 622 3,522 1.20 4,227 47%
2000 4,750 1,000 75 925 n.a. n.a. n.a.
Total
w/o 2000 $40,000 $21,359 $3,011 $18,348 n.a. $19,277 48%
Note: Columns 3–5 evaluated @ 6/30/00.
The five-year average filtered loss ratio is 48%. There seems

to be no particular trend, and there are no significant rate revi-
sions planned. So we simply take the average to be our estimate
of the 2001 loss ratio. According to Step 7, we must add on 12%
for catastrophe losses. We thus have a total expected loss ratio
of 60%.
Step 10: Estimate a “credibility” loss cost or loss cost rate from
We must reconcile the experience rating estimate of a 60%
loss ratio with our exposure rating estimate of a 65% loss ratio.
Remember that the exposure estimate includes unlimited catas-
trophe losses. The reconciliation is a process of asking questions
and judging the relative credibility of the two loss ratio estimates.
In the later discussion of treaty working cover excess pricing, we
will list some of the questions we should ask.
Let us suppose that our “credibility” estimate of the 2001
expected loss ratio is 62%.
Steps 11–13 are very intertwined. Normally, we would want
to perform Step 11 before Step 12, since Step 11 helps us quan-
tify the risk transfer on the contract with respect to the particular
contract terms. But sometimes, we have preliminary estimates of
RCR, RIXL and RTER which may later be modified. The nego-
tiations with the cedant will often send us back to Step 11 or
12, or even to earlier steps. We will present Steps 11 and 12
simultaneously.
Step 11: Estimate the probability distribution of the aggregate

reinsurance loss if desirable, and perhaps other distributions such
as for claims payment timing.
Step 12: Specify values for RCR, RIXL, and RTER

Let us start with preliminary values for Step 12. Suppose that
a review of the cedant’s expenses for the past few years indicates
an expense ratio of 33% for this business. Suppose that the cedant
will accept a ceding commission of 33%. Does this leave the

reinsurer with enough profit?
With a 33% ceding commission, the expected reinsurance pre-
mium net of commission is (1 ! RCR) " RP = 0:67 " $2,500,000
= $1,675,000. Suppose we determine that the reinsurer’s costs
for underwriting and accounting for this treaty, plus a general
overhead allocation, will be about $50,000. This translates into
an RIXL of $50,000=$1,675,000 = 3:0% (approx.). The reinsur-
ance risk pure premium with respect to a 33% ceding commis-
sion and a 3.0% reinsurer’s internal expense loading is as fol-
lows.
Formula 7.26: Reinsurance Risk Pure Premium
RPP = RP " (1 ! RCR) " (1 ! RIXL)

= $2,500,000 " 0:67 " 0:97
= $1,675,000 " 0:97
= $1,624,750
Since the reinsurer’s expected loss ratio is 62%, the expected
loss is 0:62 " $2,500,000 = $1,550,000. This leaves a profit of
$1,624,750 ! $1,550,000 = $74,750. Is this enough?
Suppose that the reinsurer generally wants RTER = 8% for the
risk transfer on this type of exposure. With RPP = $1,624,750,
the usual 8% profit margin is expected to be 0:08 " $1,624,750 =
$129,980. In addition, since there will be a statutory surplus loan
equal to the ceding commission on the unearned premium reserve
at year-end, the reinsurer should add some margin for this. Sup-
pose the historical premium data indicate that the average year-
end unearned premium reserve is about 40% of written premium.
With a 33% ceding commission, the reinsurer will be giving da
Ponte an estimated 0:33 " 0:4 " 0:25 " $10,000,000 = $330,000
of surplus relief at the end of 2001 for the 25% quota share.
Suppose that an appropriate loan rate would be 7%. Thus,
0:07 " $330,000 = $23,100 should be added to the reinsurer’s

profit margin. Adding $23,100 to the profit margin produces an
overall profit margin of $129,980 + $23,100 = $153,080. This is
$153,080 ! $74,750 = $78,330 more than the expected profit.
It is time for us to sharpen our pencil. Do we want to accept
this risk below our desired margin? Is this cedant that valuable
to us? Do we already have other business with da Ponte whose
profit makes up for this reduced margin, or will they cede such
business to us? Do we already have or can we build a long-term
profitable relationship with da Ponte and a close personal rela-
tionship with their president, Yakema Canutt? We also remember
that our estimates may have a large random error component.
What is the competitive situation?
Perhaps we can interest da Ponte in a more sophisticated con-
tractual alternative involving a variable commission rate, or a
loss and profit carryforward to successive years if we believe
that primary rates, and thus loss ratios, will improve in the next
few years. To evaluate a variable commission rate or a carry-
forward provision properly, we need an estimate of the ceded
aggregate loss distribution. So now we return to Step 11.
There are many ways to estimate the probability distribution
of the aggregate reinsurance loss. In a situation like this, where
we are not dealing with aggregate deductibles or aggregate lim-
its, it is best to use a simple method. The simplest method is
to estimate the first two moments of the aggregate loss distribu-
tion from the annual historical aggregate filtered loss data plus a
component for the catastrophe exposure. The obvious problem
with this is that in this case we have only five data points for the
filtered losses, a very small sample.
Considering the filtered loss ratios in column 8 of Table 7.25,
the mean of the policy years 1995–1999 is 48% and the standard
deviation is 6.1%. We use loss ratios instead of dollars to nor-
malize the distribution, so it can be used for policy year 2001.
Since the 6.1% estimate is derived from a small sample, we ask
whether or not it is comparable with other standard deviation

estimates from similar exposure situations. Let us assume it is.
Suppose our evaluation of the catastrophe loss potential yields
the estimated expectation of 12% mentioned previously, and an
estimated standard deviation of 13.9% (details in Appendix B).
It is appropriate to assume that the filtered loss and catastrophic
loss are independent. Under this assumption, their variances are
additive. Let us further adjust the expectation of the filtered loss
to 50% in order that the expectations add to 62%. Assuming that
the reinsurance premium RP is constant, we have the following
estimates.
Formula 7.27: First Two Central Moments of the Loss Ratio Dis-
tribution
RELC
= 62%
RP
* + * + * +
RL RLF RLC
Var = Var + Var
RP RP RP
= $(6:1%)2 + (13:9%)2 % estimate
= 2:29%
where RLF = filtered reinsurance loss
RLC = catastrophic reinsurance loss
We thus have an estimate of the standard deviation, SD[RL=RP]
= 15:13%. We approximate the loss ratio distribution with a
Gamma distribution. This is described and justified in Appendix
C, and also in Papush, Patrik and Podgaits [17]. We will later dis-
cuss more sophisticated models for approximating the aggregate
loss or loss ratio distribution. In our simple case, it is enough
to assume that the distribution of L can be represented by a
Gamma distribution whose parameters can be estimated by the
Method of Moments. We can now evaluate the ceding commis-
sion terms with this distribution, and thus estimate the reinsurer’s
profit.
Let us now finalize Step 12 before negotiations. We will use

the values for RIXL and RTER already obtained. Let us now
evaluate a sliding scale commission arrangement of the follow-
ing form in order to specify a formula for RCR. Proportional
treaties often have sliding-scale or contingent commissions. In
these cases, the reinsurer pays the cedant a provisional commis-
sion on the reinsurance gross written premium as it is transferred
to the reinsurer. At suitable dates (often quarterly), the cumula-
tive experience on the treaty (usually from the beginning if there
is a deficit carryforward, or over some period such as three years)
is reviewed. If it is profitable, the reinsurer pays the cedant an
additional commission; if it is unprofitable, the cedant returns
some of the provisional commission to the reinsurer. An exam-
ple should clarify this.
Formula 7.28: Sliding Scale Commission Arrangement

$ %
RL
RCR = PRCR ! SF " ! (1 ! PRCR ! RM)
RP
Subject to min RCR & RCR & Max RCR
where PRCR = provisional reinsurance commission rate
= 33%
SF = slide factor = 50%
RM = reinsurer’s margin = 5%
in RCR = minimum RCR = 25%
Max RCR = maximum RCR = 35%
This ceding commission formula varies the commission one
half point with each loss ratio point, with a minimum of 25%
and a maximum of 35%. Loss development is an issue to be
aware of in these types of formulas. If it is not accounted for
in the loss evaluation that determines the loss ratio, the formula
will compute too high a commission. Thus, the reinsurer will
normally be in a position of paying some commission that will
eventually be returned. This may usually be a minor point, but it

does effect the reinsurer’s cash flow, and thus economic profit,
on the contract.
If we use the Gamma distribution with the first and second
moments estimated in formula 7.27, then the expected reinsur-
ance profit becomes $108,512, as shown in Table 7.29. This is
still less than the desired profit margin of $153,080. Many rein-
surers who think that the da Ponte Insurance Company is or will
be a good client, and thus want to deal with them in the future,
will accept these terms. Or the reinsurer may request a change
in the commission terms: perhaps a reduction in PRCR together
with an increase in RM.
Please note that the table treats the Gamma as being discreet,
and displays probabilities for only selected loss ratios; the com-
putation of the average profit uses the more complete distribution
shown in Appendix C. Also note the nonsymmetry of the prob-
abilities in column 2; the skewed right-hand tail is typical of
property and casualty aggregate loss distributions.
A more sophisticated evaluation would account for cash flow
timing, e.g., when commission adjustments would be made. On
the basis of the distribution of the reinsurer’s profit displayed
in columns 2 and 4, we may specify a different target RTER.
Suppose this profit distribution is more compact than usual for
contracts of this type. This would indicate that there is less risk
than usual, thus indicating that a lesser RTER value is appropri-
ate. If the profit distribution is more spread than usual, thereby
indicating more risk, then a greater value is appropriate.
Step 13: Negotiate, reconcile opinions and estimates, alter terms

and finalize
We now have contract terms that meet our underwriting stan-
dards and largely meet our profit guidelines. We also believe
these terms will be acceptable to da Ponte Insurance Company.
Then we meet with them and explain our analysis and evaluation,
TABLE 7.29
Aggregate Loss Distribution and Calculation of
Reinsurer’s Profit
(1) (2) (3) (4)

Probability
Loss Ratio Density Reinsurance Reinsurer’s
(selected) (as percent) Commission Profit
20% 0.4% 35.0% $1,074,750

25% 3.5% 35.0% 949,750
30% 16.1% 35.0% 824,750
35% 47.2% 35.0% 699,750
40% 100.3% 35.0% 574,750
45% 166.2% 35.0% 449,750
50% 226.5% 35.0% 324,750
55% 263.3% 35.0% 199,750
60% 268.5% 34.0% 99,750
65% 245.4% 31.5% 37,250
70% 204.2% 29.0% (25,250)
75% 156.8% 26.5% (87,750)
80% 112.2% 25.0% (175,250)
85% 75.5% 25.0% (300,250)
90% 48.1% 25.0% (425,250)
95% 29.2% 25.0% (550,250)
100% 16.9% 25.0% (675,250)
105% 9.5% 25.0% (800,250)
110% 5.1% 25.0% (925,250)
115% 2.7% 25.0% (1,050,250)
120% 1.3% 25.0% (1,175,250)
125% 0.7% 25.0% (1,300,250)
130% 0.3% 25.0% (1,425,250)
135% 0.1% 25.0% (1,550,250)
140% 0.1% 25.0% (1,675,250)
145% 0.0% 25.0% (1,800,250)
150% 0.0% 25.0% (1,925,250)
$108,512
answer their questions and listen to their counter-arguments and

alternative interpretations of information and data. If necessary,
we redo the previous steps until we reach a contract structure
that satisfies both parties, if possible.
Evaluating an Ongoing Treaty

A reinsurer is in a similar evaluation situation when inter-
preting the experience on an ongoing treaty. But in addition to
evaluating the year 2001 profit potential, we must also evaluate
our cumulative experience on the treaty and compare the cu-
mulative profit to our guideline RTER (including the charge for
year-end statutory surplus loans, if any). The advantage is that
we now have our own experience data on the contract for some
years, and we also know the cedant and their people better. The
questions are similar to the previous pencil-sharpening exercise.
How much money do we think we have made, or lost? Do we
have other business with this cedant? What is the total account
bottomline profit? Is this a good relationship to continue? How
credible is our answer?
Property Surplus Share Treaties

A property surplus-share treaty is somewhat more difficult
to evaluate. Since the reinsurer does not provide coverage for
small insureds, and covers larger insureds in proportion to their
size above some fixed retention, the reinsurer must be more con-
cerned with the cedant’s pricing of larger insureds. An example
should clarify this.
Example 7.30:
' Four line first surplus not to exceed $800,000

' Maximum cedant retention = $200,000
Then the following statements are true:
' Maximum reinsurance limit per policy = $800,000
' For a policy with limit < $200,000, the reinsurer receives no
premium and pays no losses.
' For a policy with limit = $500,000, the reinsurer receives 60%
of the policy’s premium less ceding commission and brokerage
fee, and pays 60% of the policy’s losses.
' For a policy with limit = $1,000,000, the reinsurer receives
80% of the policy’s premium less ceding commission and bro-
kerage fee, and pays 80% of the policy’s losses.
' For a policy with limit = $2,000,000, the reinsurer receives
40% of the policy’s premium less ceding commission and bro-
kerage fee, and pays 40% of the policy’s losses.
It is easy to see that, given this complicated proportional struc-
ture depending upon the limit of each policy, the premium and
loss accounting for a surplus-share treaty is somewhat complex.
Despite this, surplus-share treaties are popular, because they pro-
vide more large loss protection than a quota-share, and are much
easier for the reinsurer to evaluate and price (usually only the
ceding commission and slide is the subject of negotiations) than
an excess treaty.
A surplus-share treaty is generally riskier relative to ceded
premium volume than is a simple quota-share. So the reinsurer
will charge a correspondingly higher margin for risk assumption.
Casualty Quota-Share Treaties

The pricing or evaluation of a true ground-up net retention
quota-share on casualty exposure would be similar to the pricing
of a property cover except that the reinsurer would have to be
very careful about loss development. The historical catastrophe
exposure would be asbestos and pollution clean-up, plus any
other mass tort situation that significantly distorts the cedant’s
claims data patterns. The future catastrophe exposure would be
(Fill in the blank: Tobacco? Medical implant devices?
Workers compensation claims arising from a major earthquake
during working hours? Etc.). The additional uncertainty arising
from the longer-tail claims development and trend estimation
would add risk to the rate determination, thus tending to increase
the reinsurer’s RTER. But the final evaluation of risk and neces-
sary RTER also depends upon the relative catastrophe exposure.
Working Cover Excess Treaties

A working cover is an excess layer where losses are expected.
An excess cover is usually riskier than a proportional cover. So
the reinsurer will be more mindful of prediction error and fluc-
tuation potential, and will charge a higher RTER for assuming
this risk. If losses are covered per-occurrence, then the reinsurer
is exposed by policy limits below the attachment point because
of the “clash” of losses on different policies or coverages arising
from the same occurrence.
The reinsurance premium is usually calculated via a reinsur-
ance rate times subject premium. However, for liability coverage,
it may be on an increased limits premium collected basis; this is
often the premium calculation method used for facultative auto-
matic programs. Here the total reinsurance premium is the sum of
the individually calculated reinsurance premiums for each policy.
Working cover treaties are often large enough so that many
of the risk parameters can be estimated either directly from the
exposure and claims history or by a credibility weighting of the
history with more general information. Ideally, the reinsurance
pricing consists of both an exposure rating and an experience
rating, together with a reconciliation of the two rates. We will il-
lustrate the differences from pricing either facultative certificates
or proportional treaties as we deal with an example. We will use
a casualty example. Pricing a property per-risk excess working
cover would be similar.
Example 7.31:
' da Ponte Insurance Company wants a proposal for a three-year

retrospective-rated treaty incepting Jan. 1, 2001
' All liability and workers compensation exposure
' Losses occurring coverage (accident year basis)

' Per-occurrence coverage
' ALAE added to indemnity for each claim
' Proposed attachment point AP = $300,000
' Proposed reinsurance limit RLim = $700,000
' RCR = 0%
' RBF = 0%
' Estimated 2001 subject premium PCP = $100,000,000
' Possible reinsurance premium range up to $10,000,000
We will follow the same rating steps we did for Example 7.24.
Step 1: Gather and reconcile primary exposure, expense, and rate

information segregated by major rating class groups.
Da Ponte’s casualty exposure is as follows.
TABLE 7.32
Casualty Exposure Categories
' Private passenger automobile

' Commercial automobile
' Premises/operations
' Homeowners Section II
' Special Multi-Peril Section II
' Workers compensation
' Personal umbrella
These categories should be further broken down to split the

underlying exposure at least according to the most significant
applicable increased limit tables and policy limits, or, in the case
of workers compensation, by major states and hazard groups. If
we cannot get data from da Ponte to do this, we can, as a last
resort, use default distributions of premiums by increased limit
tables and policy limits for this type of company, that we have
estimated from general information.
We should want to start with the information listed in Step
1 for Example 7.24. In addition, we assume that our underwrit-
ers have visited da Ponte and have performed an underwriting
review. We want to know about deviations from bureau man-
ual rates and average schedule and experience credits. We want
historical premiums for at least the last five-and-one-half years
1995 through June 30, 2000 plus predictions for 2001–2003. We
also want the names of contact people at da Ponte to talk with,
in particular, their pricing actuary.
Step 2: Calculate an exposure expected loss cost, PVRELC, and,

Da Ponte writes limits up to $10,000,000, but purchases fac-
ultative reinsurance for coverage in excess of $2,000,000. They
also purchase facultative coverage above $300,000 for any diffi-
cult exposures on the reinsurer’s exclusion list, and a 90% fac-
ultative automatic cover for their umbrella programs.
Treaty excess exposure rating differs from the facultative cer-
tificate excess pricing in Example 7.19 in that the reinsurer
deals with broad classes of business by increased limit table or
state/hazard group instead of individual insureds. We consider
manual rate relativities to bureau rates and/or to other companies
writing the same exposure, and evaluate da Ponte’s experience
and schedule rating plans and pricing abilities. The increased
limit factors used by the cedant for liability coverages are espe-
cially important. The same Formulas 7.2 and 7.6 can be used.
Since the coverage is per-occurrence, we must load the manual
difference rate for the clash exposure.
Since ALAE is added to each claim in order to determine
the reinsurer’s excess share, claims from some policy limits be-
low the attachment point will bleed into the excess layer. We
may have our own data describing the bivariate distribution of
indemnity and ALAE, or we may obtain such information from

ISO. Using these data, we can construct increased limits tables
where ALAE is added to loss instead of residing entirely in the
basic limits coverage.
Alternatively, and more simplistically, we can adjust the man-
ual increased limits factors to account for the addition of ALAE
to loss. A simplistic way of doing this is to assume that the
ALAE for each claim is a deterministic function of indemnity
amount for the claim, adding exactly °% to each claim value
for the range of claim sizes that are near the layer of interest.
Note that this ° factor is smaller than the overall ratio of ALAE
to ground-up indemnity loss, because much of the total ALAE
applies to small claims or claims closed with no indemnity. In
our example then, we hypothesize that when ALAE is added to
loss, every claim with indemnity greater than $300,000=(1 + °)
penetrates the layer $700,000 excess of $300,000, and that the
loss amount in the layer reaches $700,000 when the ground-up
indemnity reaches $1,000,000=(1 + °). From this you can see
how to modify standard increased limits factors to account for
ALAE added to loss. In this liability context, Formula 7.6 can
be reinterpreted with PCP as the basic limit premium and PC-
PLR as the primary company permissible basic limits loss ratio.
Given the clash exposure, suppose we believe that for this type
of coverage, an overall loss loading of ±% is sufficient to adjust
the loss cost for this layer estimated from the stand-alone poli-
cies. Then ELCF calculates the excess loss in the layer $700,000
excess of $300,000 arising from each policy limit (plus its con-
tribution to clash losses) as a percent of basic limits loss arising
from the same policy limit. The formula for ELCF evaluated at
limit, Lim, is displayed in the following formula.
Formula 7.33: Liability ELCF for ALAE Added To Indemnity Loss

ELCF(Lim) = 0
AP
if Lim &
1+°
$ ' (%
AP
ELCF(Lim) = (1 + ±) " (1 + °) " ILF(Lim) ! ILF
1+°
AP AP + RLim
if < Lim &
1+° 1+°
ELCF(Lim) = (1 + ±) " (1 + °)
$ ' ( ' (%
AP + RLim AP
" ILF ! ILF
1+° 1+°
AP + RLim
if < Lim
1+°
where AP = attachment point = $300,000
RLim = reinsurance limit = $700,000
± = clash loading = 5%
° = excess ALAE loading = 20%
Table 7.34 displays this simplistic method for a part of da
Ponte’s GL exposure using hypothetical increased limits factors
(excluding both ALAE and risk load) to calculate excess loss cost
factors. In Table 7.34, we see that policies with limits $300,000,
$500,000 and $1,000,000 and above expose the excess layer.
TABLE 7.34
Excess Loss Cost Factors with ALAE Added To
Indemnity Loss @ 20% Add-on and a Clash Loading
of 5%
(1) (2) (3)

ILF w/o risk load
Policy Limit and w/o ALAE ELCF
$ 100,000 1.0000 0
250,000 1.2386 0
300,000 1.2842 0.0575
500,000 1.4084 0.2139
833,333 1.5271 0.3635
1,000,000 or more 1.5681 0.3635
Using Formula 7.33, we calculate ELCF($300,000) = 1:05

" 1:20 " (1:2842 ! 1:2386) = 0:0575, and ELCF($1, 000, 000) =
1:05 " 1:20 " (1:5271 ! 1:2386) = 0:3635. Back up for Table
7.34 is contained in Appendix D.
Suppose da Ponte’s permissible basic limit loss ratio for this
exposure is PCPLR = 65%. Suppose our evaluation indicates that
their rates and offsets are adequate, so RCF = 1:00. Then we can
translate Table 7.34 into an exposure rate RELC, the reinsurer’s
estimate of loss cost (undiscounted) in the excess layer as shown
in Table 7.35.
TABLE 7.35
RELC (Reinsurance Expected Loss Cost (undiscounted))
(1) (2) (3) (4) (5) (6)

Estimated Estimated
Subject Basic Limit
Policy Premium Manual Loss Cost ELCF RELC
Limit Year 2000 ILF 0:65 " (2)=(3) from 6.37 (4) " (5)
Below $1,000,000 1.10 $ 590,909 0 $0

$ 300,000 (avg.)
300,000 1,000,000 1.35 481,482 0.0575 27,664
500,000 1,000,000 1.50 433,333 0.2139 92,711
1,000,000 2,000,000 1.75 724,857 0.3635 270,036
or more (avg.)
Total $5,000,000 n.a. $2,248,810 n.a. $390,411
The estimation of RELC for other categories would be similar.

For liability exposure categories where increased limit factors
are not published, use judgment to assign appropriate factors.
For workers compensation, excess loss cost factor differences
would be weighted by estimated subject premium by hazard
group, by major state grouping, or the underlying claim severity
distributions could be estimated by state, by hazard group. We
would then combine these RELC estimates with our estimates
of loss discount factors for each exposure in order to calculate
PVRELC.
A better way of estimating an exposure loss cost is to work

directly from probability models of the claim size distributions.
This directly gives us claim count and claim severity infor-
mation to use in our simple risk theoretic model for aggre-
gate loss. Suppose we know that the indemnity loss distribu-
tion underlying Table 7.34 is Pareto with b = 5,000 and q = 1:1.
Then our simple model of adding 20% ALAE to indemnity
per claim (per-occurrence) changes the indemnity Pareto dis-
tribution to another Pareto with b = 5,000 " 1:20 = 6,000 and
q = 1:1. Please note that these parameters are selected sim-
ply to make the computations easier for you to check. We
then multiply by 1 + ± = 1:05 to adjust the layer severity for
clash.4 We can then calculate excess expected claim sizes
from each policy limit. Dividing by the RELC for each limit
yields estimates of expected claim count. This is done in Table
7.36.
TABLE 7.36
Excess Expected Loss, Claim Severity, and Count
(1) (2) (3) (4)

Expected Expected
Policy Claim Size Claim Count
Limit RELC (App. E) (2)=(3)
$ 300,000 $ 27,664 $ 57,016 0.485

500,000 92,711 212,210 0.437
1,000,000 or more 270,036 360,517 0.749
Total $390,411 $233,624 1.671
Dividing RELC by the estimate of the total expected claim

count, we can back into an estimate of the total excess expected
claim size of $233,624 for this exposure. Assuming indepen-
4 One may argue that clash affects the excess claim frequency, not the excess claim
severity. The truth is that it affects both. Here, for simplicity, we only adjust the excess
claim severity.
dence of claim events across all exposures, we can also add to

obtain estimates for the expected claim (occurrence) count and
overall excess expected claim (occurrence) size.
Now we turn to experience rating.
Step 3: Gather and reconcile primary claims data segregated by
As in Example 7.24, we want the claims data segregated
as the exposure data, and we want some history of individ-
ual large claims. We usually receive information on all claims
greater than one-half the proposed attachment point, but, the
more data, the better. Suppose a claims review has been per-
formed. In our example, suppose we receive a detailed history for
each known claim larger than $100,000 occurring 1990–2000,
evaluated 12=31=90, 12=31=91, : : : , 12=31=99, 6=30=00.
Step 4: Filter the major catastrophic claims out of the claims data.
In our example, we want to identify clash claims, if possible,
and any significant mass tort claims. By separating out the clash
claims, we can estimate their frequency and size relative to non-
clash claims, and compare these statistics to values we know
from other cedants, thus enabling us to get a better estimate for
our ± loading. It should be obvious that the mass tort claims need
special treatment.
Step 5: Trend the claims data to the rating period.

As in Example 7.24, the trending should be for inflation and
for other changes in exposure (such as higher policy limits) that
might affect the loss potential. Unlike proportional coverage, we
may not skip this step. The reason is the leveraged effect of
inflation upon excess claims: a constant inflation rate increases
the aggregate loss above any attachment point faster than the
aggregate loss below, because claims grow into the excess layer,
while their value below is stopped at the attachment point. We
trend each ground-up claim value at each evaluation, including
ALAE, from year of occurrence to 2001. For example, consider

the treatment of a 1993 claim in Table 7.37.
TABLE 7.37
Trending an Accident Year 1993 Claim
(1) (2) (3) (4) (5)

Evaluation Value at Trend 2001 Level Excess
Date Evaluation Factor Value Amount
12/31/93 $0 1.59 $0 $0
12/31/94 0 1.59 0 0
12/31/95 125,000 1.59 198,750 0
12/31/96 125,000 1.59 198,750 0
12/31/97 150,000 1.59 238,500 0
12/31/98 200,000 1.59 318,000 18,000
12/31/99 200,000 1.59 318,000 18,000
6/30/00 200,000 1.59 318,000 18,000
Note the use of a single trend factor. The reasoning here is that
the trend affects claim values according to accident date (or year),
not by evaluation date. Of course a more sophisticated model for
claims inflation could be used. A delicate issue is the trending
of policy limits. If a 1993 claim on a policy with limit less than
$250,000 inflates to above $300,000 (including ALAE), would
the policy limit sold in 2001 be greater than $250,000, so to
allow this excess claim (including ALAE @ 20%)? The da Ponte
underwriter and your own marketing people may argue that the
policy limit does not change. But, over long time periods, it
would appear that the answer is that policy limits do change with
inflation. If possible, information on the da Ponte’s policy limit
distributions over time should be obtained. If this is a real issue,
you can try some sensitivity testing on the extreme alternatives:
1. The historical policy limits change with claims inflation.

2. The historical policy limits remain constant.
Then reach some judgmental compromise.
Step 6: Develop the claims data to settlement values.

We want to construct historical accident year/development
year triangles from the data produced in column 5 of Table 7.37
for each type of large claim. We would usually combine together
all claims by major line of business. Then we can use standard
methods to estimate loss development factors and apply these to
the excess claims data. We also want to compare the develop-
ment patterns estimated from da Ponte’s data to what we expect
based upon our own historical data for comparable coverages, in
order to check for reasonableness. This may be a problem, even
for filtered claims, if the historical data has insufficient credi-
bility. Consider the claim in Table 7.37: only $18,000 exceeds
the attachment point, and only at the fifth development point.
Suppose our triangle looks like Table 7.38.
TABLE 7.38
Trended Historical Claims
in the Layer $700,000 Excess of $300,000
(in $1,000’s)
Development Year
Acc. Year Age 1 Age 2 Age 3 ::: Age 9 Age 10 Age 10.5
1990 $0 $80 $254 ::: $259 $321 $321

1991 0 0 148 ::: 743 788
.. .. .. ..
. . . .
1998 57 117 236
1999 0 0
2000 0
ATA 4.236 1.573 1.076 ::: 1.239 n.a. n.a.
ATU 15.026 3.547 2.255 ::: 1.301 1.050 = tail
Smoothed 10.9% 28.7% 45.7% ::: 93.1% 95.3% 96.1%
Lags
where ATA = Age-To-Age development factor

ATU = Age-To-Ultimate development factor
Lag(t) = percentage of loss reported at time t
The tail factor of 1.05 is selected based upon general infor-

mation about development beyond ten years for this type of ex-
posure.
By switching the point of view from age-to-ultimate factors
to their inverse, time lags of claim dollar reporting, as we do
later in the Loss Reserving section, we transform the loss re-
porting view to that of a cumulative distribution function (cdf)
whose domain is [0, )). This gives us a better view of the loss
development pattern. It allows us to consider and measure the
average (expected) lag and other moments, which can then be
compared to the moments of the loss development patterns from
other exposures.
Since excess claims development is almost always extremely
chaotic, it is a good idea to employ some kind of smoothing tech-
nique, as discussed in the later loss reserving section. If properly
estimated, the smoothed factors should yield more credible loss
development estimates. They also allow us to unambiguously
evaluate the function Lag( ) at any positive time. This is handy,
since our latest data evaluation point is 6/30/00, and we want to
use these data. The smoothing introduced in the last row of Table
7.38 is based upon a Gamma distribution with mean 4 (years)
and standard deviation 3. This model is based upon the data, plus
our judgment regarding the general pattern we should expect for
loss development, the shape of the cdf. A more sophisticated
approach could estimate parameters for the Gamma distribution
directly from the development data via the method of Maximum
Likelihood Estimation by treating the data for each accident year
as a sample truncated at the last age.
Frequently, it is also useful to analyze large claim paid data,
if they are available, both to estimate excess claims payment
patterns and to supplement the ultimate estimates based only
upon the reported (incurred with no IBNR) claims used above.
It is sometimes true that the paid claims development, although
it is of longer duration, is more stable than the reported claims
development.
Occasionally, only aggregate excess claims data are avail-

able. This would be an historical accident year/development year
$700,000 excess of $300,000 aggregate loss triangle. Pricing in
a situation like this, with no specific information about large
claims, is very risky, but it is sometimes done. Technically, in
such a situation, the RTER should be higher because of the added
risk because of greater (mis-)information uncertainty. But often,
the opposite occurs because of marketing pressures.
Step 7: Estimate the catastrophic loss potential.

It should be obvious that mass tort claims need special treat-
ment. Asbestos and pollution clean-up claims, if any, distort the
historical data. As is done for property coverage, an analysis of
da Ponte’s exposures can allow us to guess some suitable loading
for future mass tort claim potential.
As we said in Step 4, we want to identify clash claims, if
possible. By separating out the clash claims, for each claim, we
add its various parts together to apply properly the attachment
point and the reinsurance limit to the occurrence loss amount. If
we cannot identify the clash claims, then our experience estimate
of RELC must include a clash loading based upon judgment for
the general type of exposure.
Step 8: Adjust the historical exposures to the rating period.

As we did in Example 7.24, we want to be sure that the his-
torical exposure (premium) data are adjusted in a manner that
makes them reasonably relevant to the rating period. The trend-
ing should be for primary rate and underwriting changes and for
other changes in exposure that might affect the loss potential.
Table 7.39 shows such a calculation.
Step 9: Estimate an experience expected loss cost, PVRELC, and,

Suppose we now have trended and developed excess losses for
all classes of da Ponte’s casualty exposure. The usual practice is
to add up the pieces to assemble an exhibit that looks like Table

7.40.
TABLE 7.40
da Ponte Insurance Company Casualty Business
(in $1,000’s)
(1) (2) (3) (4)

Trended and Estimated
Developed Loss Cost
Accident Excess Loss Rate
Year Onlevel PCP (Estimated RELC) (3) ( (2)
1992 $ 85,847 $ 3,357 3.91%
1993 87,953 4,644 5.28%
1994 89,076 6,761 7.59%
1995 92,947 5,410 5.82%
1996 94,172 4,567 4.85%
1997 95,674 3,329 3.48%
1998 98,561 4,268 4.33%
1999 99,226 6,420 6.47%
2000 49,750 1,413 2.84%
Total $793,206 $40,168 5.06%
Total $743,456 $38,755 5.21%
w/o 2000
The eight-year average loss cost rate, eliminating 2000 as be-

ing too green (undeveloped), is 5.21%. There seems to be no
particular trend from year to year. If there were, we would want
to see what its extrapolation to 2001 might be. The standard de-
viation is 1.37%. These estimates seem to be quite stable. This
is not always the case.
Alternative Estimation Methods and Some Potential

Problems
Problem: Chaotic excess development
Sometimes the loss development data for the layer $700,000
excess $300,000 for some of the categories of exposure are more
chaotic than we have seen in this example, so that we can’t esti-

mate model parameters with enough confidence. We have alter-
natives:
1. We can use our general loss development patterns for

these types of exposure, this attachment point.
2. We can experience rate a lower layer, where we may
have more stable, more credible development patterns.
For alternative 2, for example, perhaps we would drop down
to experience rate the layer $800,000 excess of $200,000, or
$300,000 excess of $200,000. We would then calculate the RELC
for $700,000 excess of $300,000 for each exposure category by
an extrapolation employing excess layer relativities derived from
claim severity curves.
Problem: Incomplete excess claims sample

If the cumulative trend factor for some exposure for some
historical accident year exceeds the ratio of the attachment point
divided by the lower bound of the large claims, then we will have
an incomplete excess sample extrapolated to the year 2001. Let
us illustrate this point with the 1993 accident year. Suppose we
were to drop down to experience rate the layer $150,000 excess
of $150,000 for 2001. Returning to the claim trending in Table
7.38, since our 1993 trend factor is 1.59, the 2001 values of all
the 1993 claims greater than $100,000 will lie above $159,000 =
1:59 " $100,000. So we are missing claims whose 2001 values
would be in the interval between $150,000 and $159,000. So
our trended claims data from 1993 are representative of the layer
$141,000 excess of $159,000, not the layer $150,000 excess of
$150,000.
Alternate Estimation Method

With the individual large claims data we have, a more sophis-
ticated methodology would have us look at excess claim count
development and excess claim size development separately. If
the data give us reasonably stable development indications, we

readily obtain claim count and claim size estimates useful for
the simple risk theoretic model for excess aggregate loss. With a
sufficient number of large claims, we can even estimate excess
claim size distribution parameters directly from the data. These
can then be compared to general parameters we have to reach
credible models for excess claim size.
Problem: Free cover

Also, an analysis that deals with individual claims may un-
cover problems buried in the simpler excess aggregate loss esti-
mation. One problem may be that of so-called “free cover.” This
arises when, in our example, the maximum trended historical
claim value is $794,826, for instance. Then the experience rate
loss cost estimate for the layer $700,000 excess of $300,000 is
the same as the experience rate loss cost estimate for the layer
$500,000 excess of $300,000. So we would be charging $0 for
the layer $200,000 excess of $800,000. It may be the case that
there are many claims whose 2001 values are close to $800,000,
and it is simply by chance that none of them exceeded $800,000.
In this case, you may simply let the experience rate loss cost
estimate stand. If this is not the case, then it looks like an insuf-
ficient rate is being calculated for coverage excess of $800,000,
and consequently a different tactic is called for. We can estimate
the loss cost for the layer $200,000 excess of $800,000 via a rel-
ative exposure rate extrapolation from the layer $500,000 excess
of $300,000. Judgment is important here, together with a review
of da Ponte’s higher policy limits exposure and a review of the
types of historical large claims that might pierce the layer excess
of $800,000.
RELC and PVRELC and PVRELC/PCP

Let us complete our estimation of the reinsurance expected
loss cost, discounted reinsurance expected loss cost, and the rein-
surance loss cost rate. Table 7.40 gives us an experience-based
estimate, RELC=PCP = 5:21%. However, this must be loaded for
whatever mass tort exposure exists, and also loaded for clash
claims if we judge that we had insufficient information on clash
claims in the claims data. A more sophisticated approach would
add in the catastrophe loss exposure rate, like we did in Example
7.24 for the property catastrophe loss.
Step 10: Estimate a “credibility” loss cost or loss cost rate from
We must also weigh the experience loss cost rate against the
exposure loss cost rate we calculated. If we have more than one
answer, and the various answers differ but cannot be further rec-
onciled, final answers for $700,000 excess of $300,000 claim
count and severity can be based upon a credibility balancing of
the separate estimates. However, all the differences should not
be ignored, but should indeed be included in your estimates of
parameter (and model) uncertainty, thus giving rise to more re-
alistic measures of variances, etc., and of risk.
Suppose we are in the simple situation, where we are only
weighing together the exposure loss cost estimate and the expe-
rience loss cost estimate. In Table 7.41 we list six considerations
for deciding how much weight to give to the exposure loss cost
estimate. You can see that the credibility of the exposure loss
cost estimate is decreased if there are problems with any of the
six items.
Likewise, in Table 7.42 we list six considerations for deciding
how much weight to give to the experience loss cost estimate.
You can see that the credibility of the experience loss cost esti-
mate is lessened by problems with any of the six items.
Appendix F has a more detailed discussion of the items in
Tables 7.41 and 7.42.
Let us assume that our credibility loss cost rate is RELC=PCP
= 5:73%.
For each exposure category, we estimate a loss discount fac-
tor. This is based upon the expected loss payment pattern for the
TABLE 7.41
Items to Consider in Determining the Credibility of the
Exposure Loss Cost Estimate
' The accuracy of the estimate of RCF, the primary rate correction factor, and
thus the accuracy of the primary expected loss cost or loss ratio
' The accuracy of the predicted distribution of subject premium by line of
business
' For excess coverage, the accuracy of the predicted distribution of subject
premium by increased limits table for liability, by state for workers
compensation, or by type of insured for property, within a line of business
' For excess coverage, the accuracy of the predicted distribution of subject
premium by policy limit within increased limits table for liability, by hazard
group for workers compensation, by amount insured for property
' For excess coverage, the accuracy of the excess loss cost factors for coverage
above the attachment point
' For excess coverage, the degree of potential exposure not contemplated by the
excess loss cost factors
TABLE 7.42
Experience Loss Cost Estimate
' The accuracy of the estimates of claims cost inflation

' The accuracy of the estimates of loss development
' The accuracy of the subject premium on-level factors
' The stability of the loss cost, or loss cost rate, over time
' The possibility of changes in the underlying exposure over time
' For excess coverage, the possibility of changes in the distribution of policy
limits over time
exposure in the layer $700,000 excess of $300,000, and upon a

selected investment yield. Most actuaries advocate using a risk-
free yield, usually U.S. Treasuries for U.S. business, for a matu-
rity approximating the average claim payment lag. Discounting
is only significant for longer tail business. So for simplicity, on
a practical basis, it’s better to use a single, constant fixed rate
for a bond maturity between five to ten years. But of course
this selection is entirely up to the reinsurer’s pricing philoso-

phy.
Let us suppose that the overall discount factor for our loss
cost rate of 5.73% is RDF = 75%, giving PVRELC=PCP = RDF
" RELC=PCP = 0:75 " 5:73% = 4:30%, or PVRELC = 4:3% "
$100,000,000 = $4,300,000.
Note that we will reverse steps 11 and 12.
Step 12: Specify values for RCR, RIXL, and RTER

Suppose our standard guidelines for this type and size of
contract and this type of exposure specify RIXL = 5% and
RTER = 15%. We can then calculate the reinsurance pure
premium of RPP = PVRELC=(1 ! RTER) = $4,300,000=0:85 =
$5,058,823, with an expected profit of RPP ! PVRELC =
$5,058,823 ! $4,300,000 = $758,823 for the risk transfer. Since
RCR = 0%, we have an indicated technical reinsurance premium
of RP = RPP=(1 ! RIXL) = $5,058,823=0:95 = $5,325,077. This
technical premium is above the maximum of $5,000,000 speci-
fied by da Ponte. Assuming that there is nothing wrong with our
technical calculations, the reinsurer has at least two options:
1. We can accept an expected reinsurance premium of

$5,000,000 at a rate of 5%, with a reduced expected
profit (assuming our internal expenses stay at $5,325,077
! $5,058,823 = $266,254) of $5,000,000 ! $4,300,000 !
$266,254 = $434,746.
2. We can offer a variable rate contract, where the reinsur-
ance rate varies according to the reinsurance loss expe-
rience, in this case a retrospectively rated contract.
Let us select the more interesting second option, especially
since, in this example da Ponte is asking for a retrospectively
rated contract. In order to construct a balanced and fair rating
plan, we need once again to estimate the distribution of the rein-
surance aggregate loss. So we proceed with step 11.
Step 11: Estimate the probability distribution of the aggregate

reinsurance loss if desirable, and perhaps other distributions such
as for claims payment timing.
We will again use a Gamma distribution approximation. But
this time, in this lower (excess) claim frequency situation, we
will obtain a better approximation of the distribution of aggre-
gate reinsurance loss using the standard risk theoretic model for
aggregate losses together with the first two moments of the claim
count and claim severity distributions.
The Standard Risk Theoretic Model for the Distribution of
Aggregate Loss
The standard model writes the aggregate loss naturally as the
sum of the individual claims (or events) as follows.
Formula 7.43: Aggregate Loss
L = X1 + X2 + * * * + XN
where L = rv (random variable) for aggregate loss
N = rv for number of claims (occurrences, events)
Xi = rv for the dollar size of the ith claim
Here N and Xi refer to the excess number of claims and the
amount of the ith claim respectively. The standard risk theoretic
model relates the distributions of L, N and the Xi ’s, as shown
in Appendix G. Under the assumption that the Xi ’s are inde-
pendent and identically distributed and also independent of N
(reasonable independence assumptions), the kth moment of L is
completely determined by the first k moments of N and the Xi ’s.
In particular, we have the following relationships.
Formula 7.44: First Two Central Moments of the Distribution of
Aggregate Loss under the Standard Risk Theoretic Model
E[L] = E[N] " E[X]
Var[L] = E[N] " E[X 2 ] + (Var[N] ! E[N]) " E[X]2
We start by assuming that E[L] = RELC = 5:73% "

$100,000,000 = $5,730,000 (undiscounted). We simplistically
assume that the excess claim sizes are independent and identi-
cally distributed and are independent of the excess claim (occur-
rence) count. In most cases this is a reasonable assumption. For
our layer $700,000 excess of $300,000, our modeling assump-
tions and results are shown in Formula 7.45, using the notation
in Appendices A and G.
Formula 7.45: da Ponte $700,000 Excess of $300,000 Aggregate

Loss Modeling Assumptions and Results
N(300) + Negative Binomial with

E[N(300)] = 24:64
Var[N(300)] = 2 " E[N(300)]
X(300) + Pareto(350, 2)
E[X(300); 700] = $232,543 censored at 700
E[X(300)2 ; 700] = 105:308 " 106
E[RL] = E[N(300)] " E[X(300); 700]
= 24:64 " $232,543
= $5,729,860
Var[RL] = E[N(300)] " E[X(300)2 ; 700]
+ $Var[N(300)] ! E[N(300)]% " E[X(300); 700]2
= E[N(300)] " $E[X(300)2 ; 700] + E[X(300); 700]2 %
= 24:64 " $105:308 " 106 + 232:543 " 106 %
= 24:64 " $159:384 " 106 %
= 3:9272 " 109
SD[RL] = $1,981,724
SD[RL]
= 0:346
E[RL]
More sophisticated modeling would more explicitly take into

account modeling and parameter risks. A good general mathe-
matical reference for “collective risk theory” is Bühlmann [9].
We will continue with our simple model here.
Let us now set up the following retrospective rating plan.
Formula 7.46: Retrospective Rate Plan
RP = RC + LF " RL
Subject to min RP & RP & Max RP
where RP = final reinsurance premium
PRP = provisional reinsurance premium
= $4,000,000
RC = reinsurance charge = $500,000
LF = loss factor = 1:00
min RP = minimum reinsurance premium
= $2,000,000
Max RP = maximum reinsurance premium
= $8,000,000
RP, PRP, RC, min RP, and Max RP can also be stated as rates
with respect to PCP. Sometimes a loss factor of 1.05 or 1.10 is
used. You can see that these are all basic parameters you can
play with in order to structure a balanced rating plan.
If we use the Gamma distribution obtained from E[RL] and
Var[RL] in formula 7.45, then the expected reinsurance profit
becomes $745,075, as shown in Table 7.47. This is close enough
to the desired profit margin of $758,823 that any reinsurer will
accept these terms.
TABLE 7.47
Aggregate Loss Distribution and Calculation of
Reinsurer’s Profit
(1) (2) (3) (4)

Probability Reinsurance
Loss Cost Rate Density Premium Reinsurer’s
(selected) (as percent) Rate Profit
0.0% 0.0% 2.00% $ 2,984

1.0% 5.2% 2.05% 2,234
2.0% 198.9% 3.10% 1,484
3.0% 914.1% 4.15% 1,109
4.0% 1765.6% 5.20% 1,109
5.0% 2120.8% 6.25% 1,109
6.0% 1886.3% 7.30% 1,109
7.0% 1363.7% 8.00% 734
8.0% 847.0% 8.00% (16)
9.0% 468.5% 8.00% (766)
10.0% 236.5% 8.00% (1,516)
11.0% 110.9% 8.00% (2,266)
12.0% 48.9% 8.00% (3,016)
13.0% 20.5% 8.00% (3,766)
14.0% 8.2% 8.00% (4,516)
15.0% 3.2% 8.00% (5,266)
16.0% 1.2% 8.00% (6,016)
17.0% 0.4% 8.00% (6,766)
18.0% 0.2% 8.00% (7,516)
19.0% 0.1% 8.00% (8,266)
20.0% 0.0% 8.00% (9,016)
Expected 5.75% $745,075
The simplified profit formula used in Table 7.47 is as follows.
Formula 7.48: Simplified Profit Formula
Reinsurer’s profit = PRP ! RIXL ! RDF " $PRP ! (RP ! RL)%

where RIXL = $266,000 (in dollars)
This simple formula assumes that the single reinsurance
claims payment, RL, and reinsurance premium adjustment, RP !
PRP, are simultaneous. A more sophisticated evaluation would

account for cash flow timing, e.g., when premium adjustments
would be made according to the premium formula timing in the
contract.
Step 13: Negotiate, reconcile opinions and estimates, alter terms

and finalize
We now have contract terms that meet our underwriting stan-
dards and also largely meet our profit guidelines. We also believe
these terms will be acceptable to da Ponte Insurance Company.
We then meet with da Ponte and explain our analysis and evalua-
tion, answer their questions and listen to their counter-arguments
and alternative interpretations of information and data. If nec-
essary, we repeat the previous steps until we reach a contract
structure that satisfies both parties, if possible.
If we cannot reach an agreement with da Ponte Insur-
ance Company on the pricing of the layer $700,000 excess
of $300,000, then the best bet is to recommend that the at-
tachment point be increased to $350,000 or $400,000. At-
tachment points should naturally increase over time in an in-
flationary environment. In this example, the expected excess
claim count of 24.64 (E[N(300)]) is fairly high for an excess
working layer. The standard deviation of about 7 (SD[N(300)])
is fairly low relative to the expectation. Perhaps now is the
time for an increase so that we and da Ponte aren’t sim-
ply trading dollars for the more predictable claims just above
$300,000.
As with facultative covers, it is clear that much of the above
can and should be programmed into an interactive computer
package for the underwriters and actuaries. And it is also ex-
tremely important to monitor the results of large treaties and
groups of treaties. The monitoring of the pricing experience and
the monitoring of IBNR emergence and the reconciliation of both
is important to the reinsurer.
Higher Exposed Excess Layer Treaties

Since there is policy limits exposure on these contracts, an
exposure rate may be determined with the same general method-
ology as for a working cover. The higher the layer, the greater
the relative significance of the workers compensation exposure,
if any, and the greater the importance of clash and other mul-
tiple limit claims’ exposure. The clash load ± in Example 7.31,
Formula 7.33 must be larger. Since losses are not “expected” for
these layers, historical loss data are sparse. And yet the layers
have loss potential, or else cedants wouldn’t buy reinsurance.
An experience loss cost rate on a contract may be calculated by
experience rating a working layer below, and using the work-
ing cover loss cost estimate as a basis for estimating the higher
exposed layer rate via extrapolation using claim size probability
distributions or non-risk-loaded increased limit factors.
Since claim frequency is lower than it is for a working layer,
the risk transfer is greater relative to premium volume. Thus,
RTER’s are greater.
Clash Treaties
Since there is no policy limit exposure, the usual exposure
rating methodology does not work, except for evaluating workers
compensation exposure. Prices for clash covers are usually set by
market conditions. The actuarial prices are largely determined by
very high RTER’s, and may or may not be close to the market-
determined prices.
For clash layer pricing, the reinsurer should keep experience
statistics on clash covers to see, in general, how often and to
what degree these covers are penetrated, and to see if the histor-
ical market-determined rates have been reasonable overall, and
also the degree to which rating cycles exist. The rates for various
clash layers should bear reasonable relationships to one another,
depending both upon the underlying exposures and upon the
distances of the attachment points from the policy limits sold.
Underwriters sometimes view clash rates with regard to a notion

of payback–the premium should cover one loss every m years
for some selected m. These kinds of underwriting-judgment rates
could be translated into judgmental expected loss cost rates plus
RIXL’s and RTER’s to put them into a consistent actuarial con-
text.
Property Catastrophe Treaties

The price for a catastrophe treaty should depend upon the
attachment point, upon the cedant’s accumulation of property
exposure in localities prone to natural catastrophes, and upon the
cedant’s net position on each policy after all other reinsurance.
Changes in the cedant’s noncatastrophe net retentions may have
a great effect upon the catastrophe exposure excess of a given
attachment point. That is, a reinsurance program can be tricky:
a small change here can have a big effect there.
The recent evolution of commercially available, large-scale
computer simulation models for natural catastrophe exposures,
especially hurricane and earthquake, has greatly increased the ac-
curacy with which these exposures can be evaluated. Of course
these models demand much more input information, such as
property premium by zip code and information about construc-
tion, etc., than was the case for the old underwriter rules-of-
thumb. The models are not yet perfect—it seems that after every
major catastrophe, the model parameters are re-adjusted to re-
flect a revised opinion regarding event frequency and severity
potential. But they are a significant step forward.
The reinsurer’s actuaries should be knowledgeable about and
be involved in the use of these models for both pricing reinsur-
ance contracts and for the measurement of the reinsurer’s ag-
gregate accumulation of catastrophe exposure. This aggregate
exposure, if not properly understood and measured, can have
the potential to blow the roof off a reinsurance company, as
was demonstrated by Hurricane Andrew in 1992. The reinsur-
ance exposure from every large property contract or portfolio of
smaller contracts must be estimated; this would be based upon

the cedants’ exposed policy limits, locality, and the reinsurance
coverage. Then the reinsurer can see where too much catastrophe
potential is accumulated, and can then better structure our own
catastrophe retrocessional program.
Aggregate Excess, or Stop Loss, Treaties

Stop loss treaties may be used to protect a cedant’s net loss
ratio. In a sense, this is the ultimate reinsurance cover for pro-
tecting a net result. Because of the magnitude of the risk transfer,
these covers are quite expensive, and often are not available un-
less written on a nontraditional basis.
For example, suppose we return briefly to the quota-share Ex-
ample 7.24. Table 7.29 displays part of the estimated probability
distribution for the loss ratio. As an alternative to the quota-share,
perhaps da Ponte Insurance Company might be interested in the
following stop loss cover.
Formula 7.49: Stop Loss Cover on da Ponte Property Net Retained

Exposure
PCP = primary company premium = $10,000,000

AP = attachment point = 70% loss ratio = $7,000,000
RLim = reinsurance limit = 30% loss ratio = $3,000,000
RP = reinsurance premium = $550,000
Suppose that the reinsurer’s expected internal expense for this
cover is about $50,000. Assuming that we use the same Gamma
distribution as before to represent the loss ratio distribution, the
profit distribution for the reinsurer is displayed in Table 7.50.
The reinsurer’s expected profit of $199,704 looks high in rela-
tion to the reinsurer’s expected loss of E[RL] = $295,230. But the
standard deviation of the reinsurer’s loss, SD[RL] = $651,000, is
extremely high compared to the expectation, and the probability
TABLE 7.50
Stop Loss Cover on da Ponte Property Exposure
(1) (2) (3) (4)

Primary Co. Reinsurer’s Reinsurer’s
Loss Ratio Loss Cost Probability Profit
70% or less $0 72.17% $ 500,000

71% 100,000 1.95% 400,000
72% 200,000 1.85% 300,000
73% 300,000 1.76% 200,000
74% 400,000 1.66% 100,000
75% 500,000 1.57% (0)
76% 600,000 1.47% (100,000)
77% 700,000 1.38% (200,000)
78% 800,000 1.29% (300,000)
79% 900,000 1.21% (400,000)
80% 1,000,000 1.12% (500,000)
81% 1,100,000 1.04% (600,000)
82% 1,200,000 0.96% (700,000)
83% 1,300,000 0.89% (800,000)
84% 1,400,000 0.82% (900,000)
85% 1,500,000 0.75% (1,000,000)
86% 1,600,000 0.69% (1,100,000)
87% 1,700,000 0.63% (1,200,000)
88% 1,800,000 0.58% (1,300,000)
89% 1,900,000 0.53% (1,400,000)
90% 2,000,000 0.48% (1,500,000)
91% 2,100,000 0.44% (1,600,000)
92% 2,200,000 0.40% (1,700,000)
93% 2,300,000 0.36% (1,800,000)
94% 2,400,000 0.32% (1,900,000)
95% 2,500,000 0.29% (2,000,000)
96% 2,600,000 0.26% (2,100,000)
97% 2,700,000 0.24% (2,200,000)
98% 2,800,000 0.21% (2,300,000)
99% 2,900,000 0.19% (2,400,000)
100% or more 3,000,000 1.37% (2,500,000)
Total $ 295,230 100% $ 199,704
that the reinsurer loses money is 19%. This is quite common

for stop loss covers. It indicates a very high risk transfer; the
reinsurer is getting a relatively small premium to cover part of
the more unpredictable tail of da Ponte’s aggregate property loss

distribution. In relation to the total premium of $10,000,000, the
reinsurer’s expected profit is only about 2%. Note the shape of
the reinsurer’s distribution of aggregate loss. It is definitely non-
Gamma. It has a positive probability of zero loss, Prob[RL =
$0] = 72:17%, and it is stopped (censored) above at 30%, or
$3,000,000, with Prob[RL = $3,000,000] = 1:37%.
If we were seriously pricing an aggregate stop loss like this,
we would construct our aggregate excess loss model more care-
fully than we did with an overall Gamma distribution. We might
use a Gamma model for the filtered claims, but build up the catas-
trophe component by modeling event frequency and severity. We
would then carefully put the two (or more) pieces together, most
likely via simulation. We would also put a lot of effort into sen-
sitivity testing our assumptions and estimates in order to see how
wrong we could be and still not get hurt too badly. For exam-
ple, we would find that if the primary expected loss ratio were
68% instead of 62%, but otherwise with the same standard de-
viation and also represented by a Gamma distribution, then the
reinsurer’s expected profit is about $0, or breakeven.
This is a simplified example. More often, to protect an ag-
gregate stop loss from catastrophe shocks, it would either cover
only noncatastrophe claims, or there would be a per-occurrence
limit.
Aggregate Excess Cover on An Excess Layer

Another form of aggregate excess treaty provides coverage
over an aggregate deductible on a per-risk or per-occurrence ex-
cess layer. Let us return to Example 7.31, covering da Ponte’s
casualty exposure $700,000 excess of $300,000. Our previous
aggregate loss modeling assumptions for this layer are listed
in Formula 7.45. Again, we will simplistically use the Gamma
model to price an aggregate deductible and limit. As discussed
above, if we were seriously pricing this coverage, we would more
carefully construct our aggregate loss model. We would also be
very careful to account for our model and parameter uncertainty

in order to get a proper spread for the aggregate loss distribu-
tion. And we would also perform a lot of sensitivity testing of our
modeling assumptions and parameters. In this excess case, with
a low claim frequency, the aggregate loss calculations would be
most likely performed via simulation or via Panjer recursion as
described in Appendix I.
Remember that for the layer $700,000 excess of $300,000 we
have the following.
Formula 7.51: $700,000 Excess of $300,000 for da Ponte’s

Casualty Exposure at a Flat Rate of 5% (Expected RP= $5,000,000)
E[RL] = $5,729,860
SD[RL] = $1,981,724
RDF = reinsurer’s loss payment discount
factor=75%
RIXL = 5% (or $266,254)
Reinsurer’s expected profit = $434,746
RTER = Reinsurer’s target economic
return (profit)
= 15%(or $758,823)
Suppose that da Ponte Insurance Company is interested in a

cover of the following form.
Formula 7.52: Aggregate Excess Cover on the Layer $700,000 Ex-

cess of $300,000 of da Ponte’s Casualty Exposure
AP = aggregate attachment point = $5,000,000

RLim = aggregate limit = $5,000,000
RP = reinsurance premium = $1,200,000
Since the loss expectation for the excess layer is $5,729,860,

with a $5,000,000 aggregate deductible, da Ponte avoids trading
dollars with the reinsurer for fairly predictable loss payments.
Keeping the premium for the deductible, da Ponte also keeps the
investment income.
Suppose that the reinsurer’s expected internal expense for this

cover is about $75,000. Also, since the reinsurer is now cover-
ing only the tail of the excess claims payments after the first
$5,000,000, let us suppose that the reinsurer’s loss discount fac-
tor decreases from 75% to 60%. Then the reinsurer’s profit dis-
tribution is displayed in Table 7.53.
Again, the reinsurer’s expected profit of $445,041 may look

high in relation to the reinsurer’s expected loss of E[RL] =
$1,133,265. But the standard deviation of the reinsurer’s loss,
SD[RL] = $1,420,620, is again very high compared to the ex-
pectation, indicating a high risk transfer. Again, the reinsurer is
getting a relatively small premium to cover part of the tail of
da Ponte’s casualty aggregate excess loss distribution. The rein-
surer’s expected profit is about the same as the $434,746 for the
previous flat-rated whole excess cover, and significantly less than
the RTER the reinsurer had wanted for that cover. The reinsurer
calculates that there is about a 3% probability that the aggre-
gate excess loss will exceed AP + RLim = $10,000,000; thus a
3% probability that da Ponte will have to pay claims in this ex-
cess tail beyond the reinsurer’s coverage. The aggregate limit of
$5,000,000 may be acceptable to da Ponte because they believe
an excess aggregate loss of $10,000,000 is impossible or, at least,
highly improbable.
As mentioned above, these kind of covers are highly sensitive

to the assumptions and estimates. The aggregate excess cover in
Formula 7.52 becomes unprofitable for the reinsurer if the ex-
pected excess loss were really 18% more than calculated, or if the
claims were paid out much faster than anticipated. When such a
cover is in place, it is important to monitor carefully the develop-
TABLE 7.53
Aggregate Excess Cover on the Layer $700,000 Excess
of $300,000 of da Ponte Insurance Company’s Casualty
Exposure
(1) (2) (3) (4)

Primary Co. Reinsurer’s Reinsurer’s
Aggregate Loss Aggregate Loss Probability Profit
$5,000,000 or less $0 39.42% $1,125,000

5,200,000 200,000 4.23% 1,005,000
5,400,000 400,000 4.17% 885,000
5,600,000 600,000 4.07% 765,000
5,800,000 800,000 3.94% 645,000
6,000,000 1,000,000 3.77% 525,000
6,200,000 1,200,000 3.59% 405,000
6,400,000 1,400,000 3.38% 285,000
6,600,000 1,600,000 3.17% 165,000
6,800,000 1,800,000 2.95% 45,000
7,000,000 2,000,000 2.73% (75,000)
7,200,000 2,200,000 2.51% (195,000)
7,400,000 2,400,000 2.29% (315,000)
7,600,000 2,600,000 2.08% (435,000)
7,800,000 2,800,000 1.88% (555,000)
8,000,000 3,000,000 1.69% (675,000)
8,200,000 3,200,000 1.52% (795,000)
8,400,000 3,400,000 1.35% (915,000)
8,600,000 3,600,000 1.20% (1,035,000)
8,800,000 3,800,000 1.06% (1,155,000)
9,000,000 4,000,000 0.94% (1,275,000)
9,200,000 4,200,000 0.82% (1,395,000)
9,400,000 4,400,000 0.72% (1,515,000)
9,600,000 4,600,000 0.63% (1,635,000)
9,800,000 4,800,000 0.55% (1,755,000)
10,000,000 or more 5,000,000 2.98% (1,875,000)
Total $1,133,265 100% $ 445,041
ment of claims payments below the aggregate excess attachment

point to see how the cover is doing. Some reinsurers have been
surprised when an aggregate cover has suddenly blown up when
claims payments reached the attachment point after many years
of no loss to the reinsurer.
Finite, or Nontraditional, Reinsurance Covers

We start with the simplest contract form.
Financial Quota-Share Treaties

The simplest example of a reinsurance cover that might be
classified as nontraditional is a financial quota-share. The Ex-
ample 7.24 quota share could be modified in various ways to
emphasize the financial aspects of the cover and decrease the
risk transfer, thus decreasing the reinsurer’s margin. For exam-
ple, in Formula 7.28, increasing the slide factor from 50% to
100% and decreasing the minimum commission from 25% to
10%, say, increases the reinsurer’s expected profit from $109,000
to $182,000 and decreases the reinsurer’s probability of losing
money from 34% to 6%. Then the reinsurer can offer a signif-
icant profit sharing arrangement for those coverage years that
run well. Introducing a loss carryforward from year to year
also makes the distribution of the multi-year loss ratio rela-
tively more compact, thus decreasing the risk transfer. Thus
the reinsurer can take a lower ultimate margin (after all adjust-
ments).
Da Ponte Insurance Company still gets surplus relief from the
provisional commission on the ceded unearned premium reserve,
and still decreases their premium-to-surplus ratio. If casualty ex-
posure were covered, the reinsurer would credit da Ponte with
some (most) of the investment income earned on the contract’s
cash balance according to some specified formula. As long as the
contract is in a profitable position, this would be returned to da
Ponte as an additional commission upon commutation or sooner.
On the other hand, there might be a penalty negative commission
if da Ponte were to cancel when the ongoing contract were in a
deficit position.
Loss Portfolio Transfers

When most insurance people think of nontraditional reinsur-
ance, they think of loss portfolio transfers. A cedant may cede
all or part of its liability as of a specified accounting date; this

may be for a line of business or territory no longer written, for
an impending sale of the company, or for other reasons. Usually,
the reinsurance premium is essentially the present value of the
transferred estimated liability, plus reinsurer’s expense, surplus
use, and risk charges. And the cedant can take reinsurance credit
for the liability ceded, thus offsetting all or part of the loss reserve
previously set up. For a U.S. cedant, this induces a surplus benefit
with respect to statutory accounting.
An example may clarify this. Suppose the da Ponte Insurance
Company has been told by its domiciliary insurance department
that it should increase loss reserves as of December 31, 2000
by 10%. With insurance department approval, da Ponte wishes
to purchase a loss portfolio cover for this additional liability.
Suppose they would like to minimize the adverse statutory sur-
plus effect as much as possible. Suppose we have the following
situation.
Example 7.54: da Ponte Loss Reserves @ 12/31/00
Carried loss reserve 12/31/00 = $300,000,000

Required additional reserve = $30,000,000
Suppose that, based upon a thorough analysis of the da
Ponte’s financial reports, historical exposure, historical reinsur-
ance program, net loss development, and claim payment distribu-
tions by line and in aggregate, we determine that the additional
$30,000,000 could easily be funded by a $15,000,000 payment.
To get to this point, besides evaluating the adequacy of their loss
reserves, we would pay careful attention to the historical claim
payment patterns and their fluctuations. Has the recent exposure
changed in such a way to cause a significant change in future
claims payments? Have there been major changes in da Ponte’s
claims department or claims processing? A common analytical
technique is to study ratios of cumulative loss payment for each
accident year divided by ultimate estimates for each category of
exposure. A simplified example is displayed in Tables 7.55 and

7.56.
TABLE 7.55
Claim Payment Development for Net GL Business:
Cumulative Paid Loss as a Ratio of Ultimate Loss
Evaluation Year (End of):

Estimated
Accident Ultimate Loss
Year (in $1,000’s) 1 2 3 ::: 9 10
1990 $60,000 0.150 0.300 0.500 ::: 0.984 1.000
1991 65,000 0.050 0.370 0.650 ::: 0.990
.. .. .. .. .. ..
. . . . . .
1997 80,000 0.100 0.380 0.550
1998 85,000 0.170 0.450
1999 90,000 0.120
1. Weighted Average 0.140 0.390 0.570 ::: 0.987 1.000
2. 3-yr Wtd Average 0.130 0.410 0.520 ::: 0.987 1.000
3. Maximum 0.170 0.450 0.650 ::: 0.990 1.000
4. Minimum 0.150 0.300 0.450 ::: 0.984 1.000
5. Trimmed Average 0.124 0.367 0.589 ::: 0.987 1.000
6. Selected “Mean” 0.124 0.367 0.589 ::: 0.987 1.000
7. Selected Extreme 0.170 0.450 0.650 ::: 1.000 1.000
The ultimate loss in the second column of Table 7.55 is our

best estimate obtained by our evaluation. In our case, the trimmed
average (weighted average of each column excluding the max-
imum and minimum) is selected as our “mean” estimate of the
claim payment distribution over time. We also select a probable
extreme value to use in our sensitivity testing.
The second column of Table 7.56, the “Estimated Liabil-
ity as % of Total,” is simply 1 ! “Mean” values from Ta-
ble 7.55; this is the expected tail left to pay for each acci-
dent year at 12/31/00. The calendar year columns display the
percent of the second column expected to be paid in each
year; these are stated as percents of the current reserve. For
TABLE 7.56
Cumulative Paid Loss as a Ratio
of Net GL Ultimate Loss: Mean Payment Distribution
Percent of Liability To Be Paid in Year:

Estimated
Acc. Liability
Year as % of Total 2000 2001 ::: 2007 2008 2009
1990 0.0% 0.0% 0.0% ::: 0.0% 0.0% 0.0%
1991 1.3% 100.0% 0.0% ::: 0.0% 0.0% 0.0%
. . . . . . . .
.. .. .. .. .. .. .. ..
1997 41.1% 38.9% 25.6% ::: 0.0% 0.0% 0.0%

1998 63.3% 35.1% 25.3% ::: 2.1% 0.0% 0.0%
1999 87.6% 27.7% 25.3% ::: 1.4% 1.5% 0.0%
Amount To Be Paid in Year (in $1,000’s):
Acc. Estimated
Year Liability 2000 2001 ::: 2007 2008 2009
1990 $0 $0 $0 ::: $0 $0 $0
1991 650 650 0 ::: 0 0 0
.. .. .. .. .. .. .. ..
. . . . . . . .
1997 36,000 14,004 9,216 ::: 0 0 0
1998 46,750 16,409 11,828 ::: 982 0 0
1999 79,200 21,938 20,038 ::: 1,109 1,188 0
Total $202,548 $70,257 $51,666 ::: $2,091 $1,188 $0
example, 27:7% = (87:6% ! 63:3%)=87:6%, 25:3% = (63:3% !

41:1%)=87:6%, 1:5% = 1:3%=87:6%, 35:1% = (63:3% ! 41:1%)
=63:3%, etc. In the lower part of Table 7.56, these percentages
are applied to our best estimate loss reserve in the second column
to get the expected claim payments in the calendar year columns.
The best estimate loss reserves are $650 = (1:000 ! 0:990)
" $650,000, $32,880 = (1:000 ! 0:550) " $80,000, etc. Then,
$12,790 = 0:277 " $32,880, etc. The Total line displays our ex-
pected claim payments by calendar year according to our “mean”
claim payment distribution. The totals include payments from
years not displayed.
We would also produce a Table 7.56 using the “extreme”

claim payment distribution from Table 7.55.
The claim payment predictions from all the covered liabili-
ties would be assembled. If a lower risk, lower margin treaty
were contemplated, greater care would be taken with the loss
discounting: the reinsurer would probably price the claim pay-
ment stream via the use of an immunizing asset portfolio. The
bond maturities would be selected to allow adequate margin for
a possible speed-up of the claim payments.
As mentioned earlier, suppose we determined that a pre-
mium payment of $15,000,000 could easily fund an additional
$30,000,000 of claims payments, since these claims payments
are out in the tail beyond the first $300,000,000 of payments. But
we wish to be more clever. To zero out the statutory surplus effect
on da Ponte, we would look for an attachment point where the
premium payment for the loss portfolio transfer would approxi-
mately match the resulting loss reserve takedown. For example,
suppose a reinsurance premium of $50,000,000 is sufficient for
a cover of $80,000,000 excess of $250,000,000. This transaction
would not change da Ponte’s beginning statutory surplus (before
reserving the additional $30,000,000). Thus, it would have zero
initial statutory effect,5 and da Ponte would be covered for the
additional $30,000,000 loss reserve.
To make the loss portfolio transfer into a finite cover, the pre-
mium would be more than necessary for a full risk cover, but
there would be substantial profit sharing if the cover ran off fa-
vorably, and the reinsurer would expect to keep a lower profit
margin. A virtual cash fund, sometimes called an experience ac-
count balance, would keep track of a fund balance, including
investment income, usually according to some prescribed calcu-
lations. If this experience account balance ended in a positive
position, or were in a positive position when the parties agreed
5 The $50,000,000 surplus benefit to da Ponte induced by this treaty would be part of
surplus under statutory accounting, but would be designated “restricted surplus.”
to commute the treaty some years hence, all or most of it would

be returned to da Ponte as profit sharing.
Funded Aggregate Excess Covers

Another example of a nontraditional reinsurance treaty is a
funded aggregate excess cover. It is clear that the aggregate ex-
cess cover of Formula 7.52 could be transformed into such a
cover by increasing the premium and introducing profit shar-
ing. One possible structure for a funded cover would be that
the reinsurer takes an initial premium of $2,000,000 instead of
$1,200,000. This increases the reinsurer’s expected profit by
$800,000 from $445,041 to $1,245,041 and decreases the rein-
surer’s probability of losing money from 26% to 11%. The rein-
surer would deduct an expense and profit margin (RIXL and
RTER combined) of perhaps only 5%, instead of the 43% pre-
viously, and allocate perhaps 90% of the calculated investment
income to the experience account balance. Further, the aggre-
gate limit at any point in time might be equal to the experi-
ence account balance plus $1,000,000, up to a maximum of
$5,000,000. Loss payments might be made only annually, to
allow the fund to grow as large as possible. As with the fi-
nancial quota-share above, there probably would be a penalty
negative commission if da Ponte cancelled in a deficit posi-
tion.
Double Triggers and Other Such Animals

With the recent advent of financial derivatives and the concept
of marketable catastrophe bonds, the area of finite reinsurance
is changing rapidly. One concept is a reinsurance coverage that
applies only if the primary insurance claims penetrate the rein-
surance layer, and also simultaneously some financial index is
above or below some designated value. Thus the name “dou-
ble trigger.” The finite reinsurance people and the investment
people are busy inventing new forms of reinsurance designed to
have very specific financial attributes. This promises to be a real
growth area.
Reinsurance Pricing: Final Comments

We have seen some examples of how standard actuarial meth-
ods and some not-so-standard actuarial methods apply to reinsur-
ance pricing. We must remember that there is no one right way
to price reinsurance. But there are many wrong ways. Common
actuarial methods should be used only to the extent they make
sense. To avoid major blunders, an underwriter/actuary must al-
ways understand as well as possible the underlying primary in-
surance exposure and must always be aware of the differences
between the reinsurance cover contemplated and that primary
exposure. The differences usually involve much less specificity
of information, longer claim report and settlement timing de-
lays, and often much lower claim frequency together with much
larger claim severity, all inducing a distinctly higher risk situ-
ation. But with this goes a glorious opportunity for actuaries
and other technically sophisticated people to use fully their the-
oretical mathematical and stochastic modeling abilities and their
statistical data analytical abilities.
In the next section, we will see how reinsurance loss reserv-
ing differs from primary insurance loss reserving, and we will
discuss some simple methods for dealing with these differences.
REINSURANCE LOSS RESERVING
General Considerations
For a reinsurance company, the loss reserve is usually the
largest uncertain number in the statement of the company’s fi-
nancial condition. To estimate a loss reserve properly, we must
study the run-off of the past business of the company. As a re-
sult of this process, we should not only be able to estimate a
loss reserve as of a certain point in time. We should also be able
to estimate historical loss ratios, loss reporting patterns, and loss
settlement patterns by year, by line, and by type of business in
enough detail to know whether or not a particular contract or
REINSURANCE LOSS RESERVING 435
business segment is unprofitable, and if so, when. This infor-

mation should also be applicable to future pricing and decision-
making. The goal is to deliver good management information
regarding the company’s historical contract portfolio, and also
deliver some indications of where the company may be going.
Reinsurance loss reserving has many of the same problems as
primary insurance loss reserving, and many of the same methods
can be used. But there are also various technical problems that
make reinsurance loss reserving somewhat more difficult. First,
we will survey some of these problems, and then examine various
methods for handling them.
Reinsurance Loss Reserving Problems

There seem to be seven major technical problems that make
reinsurance loss reserving somewhat more difficult than loss re-
serving for a primary company. These technical problems are as
follows.
Problem 1: Claim report lags to reinsurers are generally longer,

especially for casualty excess losses.
The claim report lag, the time from date of accident until first
report to the reinsurer, is exacerbated by the lengthy reporting
pipeline. A claim reported to the cedant must first be perceived
as being reportable to the reinsurer, then must filter through the
cedant’s report system to its reinsurance accounting department,
then may journey through an intermediary to the reinsurer, then
must be booked and finally appear in the reinsurer’s claim sys-
tem. The report lag may also be lengthened by an undervalu-
ation of serious claims by the cedant—for a long time an ulti-
mately serious claim may be valued below the reinsurance report-
ing threshold (usually one half of an excess contract attachment
point). This is not an indictment of primary company claims
staffs, but simply an observation that a claims person, faced with
insufficient and possibly conflicting information about a poten-
tially serious claim, may tend to reserve to “expectation,” which
is most likely interpreted by the claims person as the mode of

the probability distribution. While this modal reserving may be
sufficent for most claims with a certain probable fact pattern, it
is those few which blow up above this modal average which will
ultimately be covered by the excess reinsurer. Thus these larger
claims generally are reported later to the reinsurer than are the
smaller claims the cedant carries net.
Also, certain kinds of mass tort claims, such as for asbestosis-
related injuries, may have really extreme delays in discovery or
in reporting to the cedant, and may have dates of loss specified
finally by a court. The extreme report lags of these claims may
have a big impact on the reinsurer’s experience. Just as we saw
these time delays adding greatly to the uncertainty in reinsurance
pricing, they also add greatly to the uncertainty in reinsurance
loss reserving.
Problem 2: There is a persistent upward development of most

claim reserves.
Economic and social inflation cause this development. It may
also be caused by a tendency of claims people to reserve at modal
values, as noted in Problem 1. Also, there seems to be a ten-
dency to underreserve allocated loss adjustment expenses. Thus,
early on, the available information may indicate that a claim will
pierce the reinsurance retention, but not yet indicate the ultimate
severity.
Problem 3: Claims reporting patterns differ greatly by reinsurance

line, by type of contract and specific contract terms, by cedant, and
possibly by intermediary.
The exposure assumed by a reinsurance company can be ex-
tremely heterogeneous. This is a problem because most loss re-
serving methods require the existence of large, homogeneous
bodies of data. The estimation methods depend upon the work-
ing of the so-called law of large numbers; that is, future devel-
opment en masse will duplicate past development because of the
sheer volume of data with similar underlying exposure. Reinsur-

ers do not have this theoretical luxury, since many reinsurance
contracts are unique. And even when there exist larger aggre-
gates of similar exposure, claim frequency may be so low and
report lags so long that there is extreme fluctuation in historical
loss data. Thus, normal actuarial loss development methods may
not work very well.
As we discussed in the pricing section, a reinsurer knows
much less about the specific exposures being covered than does
a primary carrier. Also, the heterogeneity of reinsurance cover-
ages and specific contract terms creates a situation where the
actuary never has enough information and finds it difficult to
comprehend what is being covered and what is the true expo-
sure to loss. This is especially true for a reinsurer writing small
shares of brokered business.
Problem 4: Because of the heterogeneity stated in Problem 3, in-

dustry statistics are not very useful.
Every two years, the Reinsurance Association of America
(RAA) publishes a summary of casualty excess reinsurance loss
development statistics in the biannual Historical Loss Develop-
ment Study [34]. These statistics give a very concrete demon-
stration of the long report and development lags encountered by
reinsurers. However, as is noted by the RAA, the heterogeneity
of the exposure and reporting differences by company must be
considered when using the statistics for particular loss reserving
situations.
For any two reinsurers, the Annual Statement Schedule P
primary line of business exposure and loss development data
are essentially incomparable. The reason for this is that Annual
Statement lines of business do not provide a good breakdown
of reinsurance exposure into reasonably homogeneous exposure
categories useful for loss reserving. And also, most reinsurers’
loss reserves are aggregated in one line of business, 30B, excess
casualty.
Proper categorization follows the pricing categories we have

already seen, and will vary by reinsurance company according to
the types of business in which the company specializes. This is a
problem because many people who are not expert in reinsurance
insist upon evaluating a reinsurer’s loss reserves according to
Schedule P statistics. For an actuary examining a reinsurer for the
purpose of loss reserving, an appropriate exposure categorization
for the particular company may not be as apparent or as easily
accomplished as for a primary company.
Likewise, ISO loss development statistics by line are not di-
rectly applicable to reinsurance loss development without signif-
icant adjustments that may greatly increase the indicated growth.
This is so because for excess coverage, the lag in reserving or
reporting claims grows with the attachment point (see Pinto and
Gogol [32]), and also because primary company direct statistics
do not reflect the additional delays noted in Problem 1 above.
See the Reinsurance Association of America Study [33] for a
comparison of reinsurance and primary claims reporting distri-
butions. The RAA Study also has a comparison of loss develop-
ment patterns by excess attachment point.
Problem 5: The reports the reinsurer receives may be lacking some

important information.
Most proportional covers require only summary claims infor-
mation. Often the data are not even split by accident year or by
coverage year, but are reported by calendar year or by underwrit-
ing year. An underwriting year is the same as a policy year—all
premiums and claims for a contract are assigned to the effective
or renewal date of each contract. Calendar year or underwriting
year statistics are not sufficient for evaluating loss liabilities by
accident year, so various interpretations and adjustments must be
made.
Even when there is individual claims reporting, as on excess
covers, there often is insufficient information for the reinsurer’s
claims people to properly evaluate each claim without exerting
great effort in pursuing information from the cedant. This is why

it is desirable to have a professional reinsurance claims staff
even though the cedant is handling the claims. Also, reinsurance
claims people are more accustomed to handling large claims with
catastrophic injuries. Thus they are able to advise the cedant’s
staff (especially in the rehabilitation of seriously injured parties),
and sometimes reduce the ultimate payments.
For loss reserving, it is useful to have an exposure measure
against which to compare loss estimates. One possible mea-
sure is reinsurance premium by year by primary line of busi-
ness. On most contracts, losses may be coded correctly by pri-
mary line, but very often the reinsurance premium is assigned
to line according to a percentage breakdown estimate made at
the beginning of the contract and based upon the distribution
of subject premium by line. To the degree that these percent-
ages do not accurately reflect the reinsurer’s loss exposure by
primary line, any comparisons of premiums and losses by line
may be distorted. This adds to the difficulties noted in Prob-
lem 4.
For most treaties, premiums and losses are reported quarterly
in arrears; they may not be reported (and paid) until some time
in the following quarter. Thus there is an added IBNR exposure
for both premiums and losses. The actuary must remember that,
at year-end, the latest year premiums may be incomplete, so they
may not be a complete measure of latest year exposure.
Problem 6: Because of the heterogeneity in coverage and reporting

requirements, reinsurers often have data coding and IT systems
problems.
All reinsurers have management information systems prob-
lems. The business has grown in size and complexity faster, and
expectations regarding the necessary level of data detail have
also grown faster, than the ability of reinsurers’ data systems to
handle and produce the reports requested by marketing, under-
writing, claims, accounting, and actuarial staffs. This problem
may be endemic to the insurance business, but it is even more

true for reinsurance.
Problem 7: The size of an adequate loss reserve compared to sur-

plus is greater for a reinsurer.
This is not a purely technical problem. It is more a manage-
ment problem, and many reinsurance companies have stumbled
over it. Problems 1–6 act to increase the size of an adequate loss
reserve and also make it more uncertain. Thus, it is difficult for
the actuary to overcome the disbelief on the part of management
and marketing people, and convince them to allocate adequate
resources for loss liabilities. Eventually, claims emerging on old
exposure overwhelms this disbelief, at least for those who listen.
A cynic might say that many reinsurance managers change jobs
often enough to stay ahead of their IBNR. Start-up operations in
particular have this problem—if there is no concrete claims run-
off experience to point to, why believe a “doom-saying” actuary?
So What?
These seven problems imply that uncertainty in measurement
and its accompanying financial risk are large factors in reinsur-
ance loss reserving. This became even more important after the
U.S. Tax Reform Act of 1986 required the discounting of loss re-
serves for income tax purposes. This discounting eliminated the
implicit margin for adverse deviation that had been built into pre-
vious insurance loss reserves simply by not discounting. Insurers
lost this implicit risk buffer. Since this buffer then flowed into
profits and thus was taxed sooner, assets decreased. This clearly
increased insurance companies’ risk level. The effect upon rein-
surers was even greater.
Components of a Reinsurer’s Loss Reserve

The six general components of a reinsurer’s statutory loss
reserve are as follows.
Component 1: Case reserves reported by the ceding companies

These may be individual case reports or may be reported in

bulk, depending upon the loss reporting requirements of each
individual contract. Most excess contracts require individual case
reports, while most proportional contracts allow summary loss
reporting.
Component 2: Reinsurer additional reserves on individual claims

The reinsurer’s claims department usually reviews individual
case reserve reports and specifies additional case reserves (ACR)
on individual claims as necessary. Additional case reserves may
vary considerably by contract and by cedant.
Component 3: Actuarial estimate of future development on Com-

ponents (1) and (2)
The future development on known case reserves in total is
sometimes known is IBNER, Incurred (and reported) But Not
Enough Reserved.
Component 4: Actuarial estimate of pure IBNR

Most actuaries would prefer that separate estimates be made
for Components (3) and (4), the estimate of pure IBNR, Incurred
But Not Reported. However, because of limitations in their data
systems, in practice most reinsurers combine the estimates of
Components (3) and (4). Depending upon the reinsurer’s mix of
business, these together may amount to more than half the total
loss reserve.
Unless otherwise noted, the term IBNR in this chapter stands
for the sum of IBNER and pure IBNR.
Component 5: Discount for future investment income

Insurance companies are allowed to take statutory account-
ing credit for future investment income on the assets support-
ing workers compensation permanent total cases, automobile PIP
annuity claims and medical professional liability claims. Some
companies do discount these claims reserves, and some don’t.

And, of course, as mentioned above, the U.S. Tax Reform Act
of 1986 requires discounting of loss reserves for income tax pur-
poses.
Component 6: Risk load

The last component of a loss reserve should be the risk load-
ing or adverse deviation loading necessary to keep the reserve
at a suitably conservative level, so as not to allow very uncer-
tain income to flow into profits too quickly. Some loss reserving
professionals prefer to build this into the reserve implicitly by
employing conservative assumptions and methodologies. How-
ever, many actuaries would prefer to see it estimated and ac-
counted for explicitly. Because of the long-tailed nature of much
of their exposure and its heterogeneity and the uncertainty of
their statistics, this component is theoretically more important
for reinsurers.
A General Procedure
The four steps involved in a reinsurance loss reserving
methodology are as follows.
Step 1: Partition the reinsurance portfolio into reasonably homo-
geneous exposure groups that are relatively consistent over time
with respect to mix of business (exposures).
It is obviously important to segregate the contracts and loss
exposure into categories of business on the basis of loss devel-
opment potential. Combining loss data from nonhomogeneous
exposures whose mix has changed over time can increase mea-
surement error rather than decrease it.
Reasonably homogeneous exposure categories for reinsurance
loss reserving have been discussed in the actuarial literature and
follow closely the categories used for pricing. Table 7.57 lists
various important variables for partitioning a reinsurance port-
folio. All affect the pattern of claim report lags to the reinsurer
and the development of individual case amounts. The listing is

meant to be in approximate priority order.
TABLE 7.57
Important Variables for Partitioning a Reinsurance
Portfolio into Reasonably Homogeneous Exposure
Categories
' Line of business: property, casualty, bonding, ocean marine, etc.

' Type of contract: facultative, treaty, finite (or “financial”)
' Type of reinsurance cover: quota share, surplus share, excess per-risk, excess
per-occurrence, aggregate excess, catastrophe, loss portfolio transfer, etc.
' Primary line of business—for casualty
' Attachment point—for casualty
' Contract terms: flat-rated, retro-rated, sunset clause, share of loss adjustment
expense, claims-made or occurrence coverage, etc.
' Type of cedant: small, large, or E&S company
' Intermediary
Obviously, a partition by all eight variables would split a con-

tract portfolio into numerous pieces, many with too little cred-
ibility. However, after partitioning by the first five variables, it
may be desirable to recognize the effects of some of the other
variables on certain classes of business. For example, whether or
not a casualty contract covers on an occurrence or claims-made
basis is obviously important. The RAA now requests loss devel-
opment data for its biannual study to be segregated by specified
attachment point ranges. These ranges are a useful guide to any
reinsurer wishing to so classify their own claims data and expo-
sure.
Each reinsurer’s portfolio is unique and extremely heteroge-
neous. In order to determine a suitable partition of exposure for
reserving and results analysis, we must depend greatly upon the
knowledge and expertise of the people writing and underwriting
the exposures, the people examining individual claim reports,
and the people processing data from the cedants. Their knowl-
edge, together with elementary data analysis (look at simple loss
development statistics), help the actuary understand which of the

variables are most important.
One possible first-cut partition of assumed reinsurance expo-
sure is shown in Table 7.58. Remember that there is no such
thing as a “typical” reinsurance company.
TABLE 7.58
Example of Major Exposure Categories for a
Reinsurance Company
' Treaty casualty excess

' Treaty casualty proportional
' Treaty property excess
' Treaty property proportional
' Treaty property catastrophe
' Facultative casualty
' Facultative property
' Surety
' Fidelity
' Ocean marine
' Inland marine
' Construction risks
' Aviation
' Finite, or nontraditional, reinsurance
' Miscellaneous special contracts, pools and associations
' Asbestos, pollution, and other health hazard or mass tort claims
The last item in Table 7.58 is not really an exposure category.

But anyone in their right mind would advocate that these types of
claims should be treated separately for any insurance company,
especially a reinsurance company.
Within the major categories, the exposure should be further
refined into treaty and facultative, if not already specified. Also,
all significant excess exposure should be further segregated by
type of retention (per-occurrence excess vs. aggregate excess).
Treaty casualty excess exposure should be further segregated
by attachment point range and by primary line (automobile
liability, general liability, workers compensation, medical profes-

sional liability). Each of these categories would be expected to
have distinctly different lags for claims reported to the reinsurer.
Categories for treaty casualty proportional business would be
similar. As we have discussed, some contracts classified as pro-
portional are not shares of first dollar primary layers, but rather
shares of higher excess layers. Thus, whether the exposure is
ground-up or excess is an important classification variable.
Loss reserving categories for facultative casualty would cer-
tainly separate out automatic primary programs (pro rata share of
ground-up exposure) and automatic nonprimary programs (ex-
cess). Certificate exposure should be split by attachment point
range, if possible, but at least into buffer versus umbrella layers,
and then further by primary line.
Likewise for property and other exposures, the loss reserving
categories should correspond closely to the pricing categories.
It will be convenient to discuss steps 2 and 3 together.
Step 2: Analyze the historical development patterns. If possible,

consider individual case reserve development and the emergence
of IBNR claims separately.
Step 3: Estimate the future development. If possible, estimate the

bulk reserves for IBNER and pure IBNR separately.
In our discussion, we will only deal with aggregate claim
dollar development of both IBNER and pure IBNR combined. For
certain exposures, especially longer-tail lines, it is often a good
idea to deal with them separately. Some techniques for doing so
are discussed in the chapter on loss reserving.
For suitably homogeneous categories of exposure, expected
reinsurance claim development patterns are very stable. How-
ever, because of the extreme variability in the year-to-year data,
these patterns should be studied using claims data over long time
periods, as long as the expected patterns are reasonably stable

from year-to-year. Usually the longer the time period, the better,
in order to obtain accurate estimates of the underlying pattern.
This is usually a large, time-consuming analytical job. Because
of year-end time pressures, it is almost always a good idea to per-
form the in-depth analysis of claim development patterns during
the third or fourth quarter. The idea is to construct our models
and estimate most of the parameters before we get to the year-
end crunch. Then, at year-end, and at the end of each quarter,
we simply apply our parameterized models to the year-end or
quarter-end claims and exposure data to estimate our IBNR (re-
member, here we mean both IBNER and pure IBNR).
Claim Report and Payment Lags

For analyzing and understanding reinsurance claims develop-
ment patterns, it is useful to consider the inverse of the usual
chainladder age-to-ultimate development factors—we call the
factor inverses “lags.” This view produces a time lag curve,
y = Rlag(t), where t measures time, like that shown in Graph
7.59.
As the age goes from 0 years to 10 years, the lag goes from
Rlag(0) = 0% to Rlag(10) = 99:9%. The graph looks like a prob-
ability cumulative distribution function (cdf), and, with a bit of
imagination, can be interpreted as one. Rlag(t) can be read as the
probability that any particular claims dollar will be reported to
the reinsurer by time t. This view of the claims reporting pro-
cess allows us to compute statistics, such as the expected value
(in years), by which we can compare one claim report pattern
with another. This view also helps us fit smooth curves to the of-
ten chaotic tails of claims development data, and compute values
at intermediate points. The claims report lag in Graph 7.59 is a
Gamma distribution with mean 2.5 years and standard deviation
1.5 years.
The same lag idea applies to claim payment patterns.
GRAPH 7.59
Claim Report Lag Graph
It is convenient to discuss the methods to be used for the

historical analyses and the estimation methods for the different
exposure categories according to the lengths of the average claim
dollar report lags.
Methods for Short-tailed Exposure Categories

As is generally true, the best methodologies to use are those
that provide reasonable accuracy for least effort and cost. For
short-tailed lines of business, such as most property coverage
exposure, losses are reported and settled quickly, so loss liabili-
ties are relatively small and run off very quickly. Thus, elaborate
loss development estimation machinery is unnecessary.
Reinsurance categories of business that are usually short-
tailed are listed in Table 7.60. But, as with any statement about
reinsurance, be careful of exceptions.
TABLE 7.60
Reinsurance Categories That Are Usually
Short-tailed (with Respect to Claim Reporting and
Development)
Category Comments,
Treaty property proportional Be wary of recent catastrophes

Treaty property catastrophe Be wary of recent catastrophes
Treaty property excess Possibly exclude high layers, and be wary of
recent catastrophes
Facultative property Exclude construction risks, and be wary of recent
catastrophes
Fidelity proportional
,
Exclude all international exposure, if possible, since there may be significant reporting delays.
Estimation Methods
Many reinsurers reserve property business by setting IBNR
equal to some percentage of the latest-year earned premium. This
is sometimes a reasonable method for non-major-catastrophe
“filtered” claims, as defined in the pricing section. It is a good
idea to consider major storms and other major catastrophes sep-
arately. A recent catastrophe will cause real IBNR liability to far
exceed the normal reserve. Claims from major catastrophes may
not be fully reported and finalized for years, even on proportional
covers.
Another simple method used for short-tailed exposure is to
reserve up to a selected loss ratio for new lines of business or for
other situations where the reinsurer has few or no loss statistics.
For short-tailed exposure, provided the selected loss ratio bears
some reasonable relationship to past years’ experience and pro-

vided that it is larger than that computed from already-reported
claims, this is a reasonable method for filtered (noncatastrophic)
claims.
For some proportional treaties, summary loss reporting may
assign claims by underwriting year, according to inception or re-
newal date of the reinsurance treaty, instead of by accident year.
If the reinsurer’s claims accounting staff records the reported
claims likewise, the loss statistics for each false “accident” year
may show great development because of future occurring acci-
dents. In this situation, a more accurate loss development picture
and estimation of IBNR can be obtained by assigning these “acci-
dent” year losses to approximate true accident year by percentage
estimates based upon the underwriting year inception date and
the general report lag for the type of exposure. Summary claims
reported on a calendar (accounting) year basis can likewise be
assigned to accident year by percentage estimates, if necessary.
For short-tailed lines for which the IBNR is estimated as a per-
centage of premium or reserved up to a selected loss ratio, these
re-assignments are unnecessary.
Methods for Medium-Tailed Exposure Categories

Let us consider any exposure for which claims are almost
completely settled within five years and with average aggregate
dollar claim report lag of one to two years to be medium-tailed
for this discussion. Reinsurance categories of business that are
usually medium-tailed are listed in Table 7.61.
Even if a property claim is known almost immediately, its ul-
timate value may not be. Thus it may take longer to penetrate a
higher per-risk excess attachment point. This happens more of-
ten if time element coverage is involved. The discovery period
for construction risk covers may extend years beyond the con-
tract (loss occurrence) period. So for both these exposures, claim
report lags may be significantly longer than normal for property
business.
TABLE 7.61
Reinsurance Categories That Are Usually
Medium-tailed (with Respect to Claim Reporting and
Development)
Category Comments
Treaty property excess higher If it is possible to separate these from working

layers layers
Construction risks If it is possible to separate these from other
property exposure
Surety
Fidelity excess
Ocean marine
Inland marine
International property
Non-casualty aggregate excess Lags are longer than for the underlying exposure
For surety exposure, it is usually a good idea to consider losses

gross of salvage and, separately, salvage recoveries. The gross
losses are reported fairly quickly, but the salvage recoveries have
a longer tail. It is instructive to consider the ratio of salvage to
gross loss for mature years. This ratio is fairly stable and may
help explain predictions for recent coverage years as long as the
underwriters can predict how the salvage ratio may have slowly
changed over time.
Estimation Methods
A useful IBNR estimation method for medium-tailed lines of
business is to use the standard American casualty actuarial Chain-
ladder (CL) Method of age-to-age factors calculated from cumu-
lative aggregate incurred loss triangles, with or without ACRs.
This is described fully in the chapter on loss reserving. If ac-
cident year data exist, this is good methodology. An advantage
of this method is that it strongly correlates future development
both with an overall lag pattern and with the claims reported
for each accident year. A major disadvantage is simply that the
IBNR is so heavily correlated with reported claims that, at least
for longer-tailed lines, the reported, very random nose wags the
extremely large tail estimate for recent, immature years.
It is sometimes true that paid loss development is more stable
than reported loss development. If so, then a chainladder estimate
of ultimate loss by accident year may be obtained using paid lags.
Of course, the problem is that the estimation error may be even
greater for immature, recent accident years than it is for reported
loss chainladder estimation.
Methods for Long-Tailed Exposure Categories

Just as for pricing, the real problem in loss reserving is long-
tailed exposure, especially excess casualty reinsurance. These are
the exposures for which the average aggregate claims dollar re-
port lag is over two years and whose claims are not settled for
many years. Reinsurance categories of business that are usually
long-tailed are listed in Table 7.62.
TABLE 7.62
Reinsurance Categories That Are Usually Long-Tailed
(with Respect to Claim Reporting and Development)
Category Comments
Treaty casualty excess Includes the longest lags except for the
APH claims listed below
Treaty casualty proportional Some of this exposure may possibly be
medium-tailed
Facultative casualty
Casualty aggregate excess Lags are longer than for the underlying
exposure
Asbestos, pollution and other health May be the longest of the long tails
hazard and mass tort claims
For most reinsurance companies, this is where most of the

loss reserve lies, and almost all the IBNR. So it is important for
us to spend most of our analytical and estimation effort here. The
first step is to separate these exposures into finer, more homoge-

neous categories. This is, of course, an iterative process. We de-
pend upon our company’s marketing, underwriting, claims, and
accounting personnel for the first-stage categorization. Further
refinements will then depend upon our hypothesis testing and
upon our investigation of various comments from the marketing
and underwriting people as they receive from us the estimated
IBNR by major contract or category based upon the latest catego-
rization. Some of the larger, more important contracts are looked
upon at least partially on a standalone basis.
Asbestos, Pollution, Other Health Hazard and Other Mass Tort

Exposures
We separate out claims arising from asbestosis, pollution
clean-up, other health hazard (sometimes collectively known as
“APH”) and other mass tort situations for special consideration.
Because of the catastrophic significance of these types of claims
(nothing for many years, then suddenly, gigantic totals), they
would drastically distort the development statistics if left in the
remaining filtered (noncatastrophic) claims data. Also, it is un-
likely that the usual actuarial loss development techniques, if
used blindly, would yield reasonable answers for these types of
claims.
In the past few years, various models have been developed
for estimating asbestos and pollution clean-up liabilities. The
large actuarial consulting firms have been in the forefront of
this development. The models are extremely complex and re-
quire very detailed input regarding coverages exposed to these
types of claims. Unless we, the reinsurer, have a very large, long,
stable and fairly complete claims development history, we judge
that it is better that we and our APH claims specialists work with
one of the actuarial consultants to develop our liability estimates.
Beyond asbestos and pollution clean-up, the question of
which exposures or claims should be treated separately or spe-
cially is difficult. We should discuss this thoroughly with our
claims and underwriting staff. For example, it is desirable to treat

claims-made coverages separately from occurrence coverages, if
possible. Also, it should be clear that claims from commuted con-
tracts should be excluded from the usual triangles, since their
development is artificially truncated (cut off), and thus would
distort the noncommuted complete development patterns.
Estimation Methods
The standard CL Method is sometimes used for long-tail ex-
posures. But, the problem is that for very long-tailed lags, the
resulting IBNR estimate for recent, green years is extremely vari-
able, depending upon the few reported or paid claims to date.
An alternative estimation method is the Bornheutter–Ferguson
(BF) Method (Bornheutter and Ferguson [25]), which is dis-
cussed in the chapter on loss reserving. This method uses a se-
lected loss ratio for each coverage year and an aggregate dollar
report lag pattern specifying the percentage of ultimate aggre-
gate loss expected to be reported as of any evaluation date. An
advantage of this method is that it correlates future development
for each year with an exposure measure equal to the reinsurance
premium multiplied by a selected loss ratio. Disadvantages with
the BF IBNR estimate are:
1. It is very dependent upon the selected loss ratio.

2. The estimate for each accident year does not reflect the
particular to-date reported losses for that year, unless the
selected loss ratio is chosen with this in mind.
Since the loss ratio for a given accident year is strongly cor-
related with the place of that year in the reinsurance profitability
cycle, so is the to-date reported loss. It would seem to be desir-
able to use this fact in the IBNR estimate. As noted before, the
reinsurance profitability cycles are more extreme than primary
insurance cycles. Thus, when using the BF Method, one must
select the accident year loss ratios carefully.
Stanard–Bühlmann (Cape Cod) Method

An estimation method that overcomes some of the prob-
lems with the CL and BF Methods was independently derived
by James Stanard (described in Patrik [31] and Weissner [36])
and by Hans Bühlmann (internal Swiss Re publications). This
Stanard–Bühlmann (SB) Method is known to European actuaries
as the Cape Cod Method because Hans Bühlmann first proposed
it at a meeting on Cape Cod. As with the CL and BF Methods,
this method uses an aggregate reported claim dollar lag pattern,
which may or may not be estimated via the CL Method or via
some other method. The key innovation of the SB Method is that
the ultimate expected loss ratio for all years combined is esti-
mated from the overall reported claims experience, instead of
being selected judgmentally, as in the BF Method. A problem
with the SB Method is that the IBNR by year is highly dependent
upon the rate-level adjusted premium by year. The user must
adjust each year’s premium to reflect the rate-level cycle on a
relative basis. But this is also a problem with the BF Method.
A simple example will help explain the SB Method.
For a given exposure category, assume that the yearly earned
risk pure premiums (net of reinsurance commissions, brokerage
fees and internal expenses) can be adjusted to remove any sus-
pected rate-level differences by year. Thus, we believe that each
adjusted year has the same expected loss ratio. In primary in-
surance terminology, assume that the risk pure premiums have
been put on-level. This adjustment is difficult and uncertain, but
must be done if we are to have a reasonably consistent exposure
base. Netting out the commissions and brokerage fees is usually
easy, since these are separately coded by contract in our account-
ing database. Netting out the internal expenses is more difficult,
since these are strictly internal numbers, not usually coded to
contract. If it is too difficult or impossible to net out the inter-
nal expense, we need not bother, since all we need is a rela-
tive exposure base, not an absolute one. We study primary busi-
ness rate levels by considering industrywide Schedule P line loss
ratios by year, and work with our underwriters to adjust these rate
relativities for our own reinsurance exposure categories. Let ELR
represent the unknown expected loss ratio to adjusted earned risk
pure premium.
Suppose that Table 7.63 displays the current experience for
this category. For clarity, to deal with an example with only five
years, we use a report lag that reaches 100% at the end of the
sixth year.
TABLE 7.63
Stanard–Bühlmann (Cape Cod) Method
Data as of 12/31/00 (in 1,000’s)
(1) (2) (3) (4) (5) (6)

Calendar/ Earned Aggregate Aggregate “Used-Up”
Accident Risk Pure Adjusted Reported Loss Premium
Year Premium Premium Loss Report Lag (3) " (5)
1996 $ 6,000 $ 8,000 $ 7,000 95% $ 7,600

1997 7,000 7,000 5,000 85% 5,590
1998 8,000 6,000 3,000 70% 4,200
1999 9,000 7,000 2,000 50% 3,500
2000 10,000 10,000 4,000 30% 3,000
Total $40,000 $38,000 $21,000 n.a. $24,250
We will explain column 6 of Table 7.63 in the following dis-

cussion.
The SB ELR and IBNR estimates are given by Formula 7.64.
Formula 7.64: Stanard–Bühlmann (Cape Cod) ELR and IBNR Es-

timates
§$RRL(k)%
SBELR =
§$ARPP(k) " Rlag(k)%
SBIBNR(k) = SBELR " ARPP(k) " (1 ! Rlag(k))
where SBELR = Stanard–Bühlmann estimate of the ELR
SBIBNR(k) = SB IBNR estimate, year k
RRL(k) = reported reinsurance loss, year k

ARPP(k) = adjusted risk pure premium, year k
Rlag(k) = aggregate claim dollar report lag, year k
Some SB practitioners call the term ARPP(k) " Rlag(k), the

used-up premium for year k. You can see that it is that fraction
of premium corresponding to the percent of ultimate claims ex-
pected to be reported as of the evaluation date. It is the premium
earned according to the expected reported claims. Thus SBELR
is the loss ratio of reported loss divided by used-up premium. So
how do we get to these equations?
First of all, simply define SBIBNR(k) to be SBELR " ARPP(k)

" (1 ! Rlag(k)). This definition is the same as in the BF Method.
The only difference is that the SB Method will tell us how to
estimate SBELR. Since the total IBNR is the sum of the IBNR
from each year, then SBIBNR can be written as follows.
SBIBNR = §SBIBNR(k)
= §$SBELR " ARPP(k) " (1 ! Rlag(k))% by definition
= SBELR " §$ARPP(k) " (1 ! Rlag(k))%
Then, since a loss ratio is simply losses divided by premium,

we may write SBELR as follows.
RRL + SBIBNR
SBELR =
ARPP
RRL + SBELR " §$ARPP(k) " (1 ! Rlag(k))%
=
ARPP
Or
SBELR " ARPP = RRL + SBELR " §$ARPP(k) " (1 ! Rlag(k))%

= RRL + SBELR " ARPP
! SBELR " §$ARPP(k) " Rlag(k)%
Or, subtracting SBELR " ARPP from both sides,
SBELR " §$ARPP(k) " Rlag(k)% = RRL

Or
RRL
SBELR =
§$ARPP(k) " Rlag(k)%
= $21,000=($8,000 " 0:95 + $7,000 " 0:85
+ $6,000 " 0:70 + $7,000 " 0:50
+ $10,000 " 0:30)
$21,000
=
$24,250
= 0:866
Table 7.65 compares IBNR and estimated ultimate loss ratios
for the CL and SB Methods, using the information in Table 7.63
(remember to use the adjusted risk pure premium, ARPP, not
the original premium, in the SB calculation). The BF Method
cannot be compared, since the BF loss ratios are not estimated
by formula.
TABLE 7.65
Comparison of Chainladder and Stanard–Bühlmann
Methods
(1) (2) (3) (4) (5) (6)

Cal/Acc. Earned Risk CL CL SB SB
Year Pure Premium IBNR Loss Ratio IBNR Loss Ratio
1996 $ 6,000 $ 368 123% $ 346 122%
1997 7,000 882 84% 909 84%
1998 8,000 1,286 54% 1,589 57%
1999 9,000 2,000 44% 3,031 56%
2000 10,000 9,333 133% 6,062 100%
Total $40,000 $13,870 87% $11,908 82%
Look at the differences in the 1999 and 2000 loss ratios. As

long as the rate relativity adjustments to yearly earned risk pure
premium are reasonably accurate, the estimates for recent, im-
mature years and the overall answer are more accurate with the
SB Method. It is easy to see that the above example would be
even more vivid if a real longer-tailed exposure were used.
Credibility IBNR Estimates

For situations where we don’t have complete confidence in the
year-to-year rate-level premium adjustments, we can calculate a
“credibility” IBNR estimate by weighing together the CL and SB
estimates. Intuitively, we would want to give more weight to the
SB estimate for recent, green years, where the CL estimate has a
high level of variance, and more weight to the CL estimate for
older years, where we trust the cumulative rate-level adjustments
less. One way of doing this is to use some simple, monotonic
function of the report lag as the credibility weight for the CL
estimate. A simple linear function that yields such a credibility
estmate as follows.
Formula 7.66: A Simple “Credibility” IBNR Estimate
Cred IBNR(k) = Z(k) " CLIBNR(k) + (1 ! Z(k)) " SBIBNR(k)

where Z(k) = CF " Rlag(k)
and 0 & CF = credibility factor & 1
Rlag(k) = report lag for year k
Cred IBNR(k) = credibility IBNR for year k
CLIBNR(k) = CL IBNR for year k
SBIBNR(k) = SB IBNR for year k
Using the information in Tables 7.63 and 7.65, we can cal-
culate a credibility IBNR, with CF = 0:5 for example, in Table
7.67.
TABLE 7.67
Calculation of ‘‘Credibility’’ IBNR
(1) (2) (3) (4) (5) (6)

Claim
Cal/Acc. Earned Risk Report “Credibility”
Year Pure Premium Lag CL IBNR SB IBNR IBNR
1996 $ 6,000 95% $ 368 $ 346 $ 356

1997 7,000 85% 882 909 898
1998 8,000 70% 1,286 1,589 1,483
1999 9,000 50% 2,000 3,031 2,773
2000 10,000 30% 9,333 6,062 6,553
Total $40,000 n.a. $13,870 $11,908 $12,063
For example, in Table 7.67, $356 = (0:5) " (0:95) " $368 +
(1 ! (0:5) " (0:95)) " $346.
If we were dealing with a real live long-tail example, we

would see many more accident years worth of data. Depend-
ing upon the length of the claim report and payment lags, we
would probably want to be looking at 25 or more years of claims
data. We would talk with our marketing people, underwriters, and
claims people to see if there are any special contracts, exposures,
or types of claims that should be treated separately. We would
want to know of any particularly large individual claims which
should be censored or otherwise dealt with so as not to have an
undue random impact upon either the estimation of the claims
report and payment lags or upon the IBNR estimation.
Other “Credibility” Procedures

Some reinsurance actuaries also weigh together the IBNR es-
timates obtained separately from reported claims data and paid
claims data. A problem with reported claims data is that they
include case reserves set up according to the judgments of many
claims people, perhaps varying over time. Thus they may lack
consistency over time. If we have paid claims data over enough
accident and payment years to capture the long tail, and we be-
lieve the data to have a fairly consistent expected payment pattern
over the accident years, these data may exhibit more stability than
the reported claims data. Thus they may also be useful for es-
timating IBNR liabilities. The weighing can again be intuitively
simplistic. Perhaps we might decide to use the relative claim re-
port and payment lags for each year as weights.
Some reinsurance actuaries also want to use the information
inherent in the pricing of the exposure. If actuarial pricing tech-
niques similar to those discussed in the pricing section are used,
we automatically have estimates of ELR with respect to risk pure
premium for each contract. We may monitor their accuracy and
average them over suitable exposure categories. We may call
these average pricing ELR’s our a priori ELR estimates and use
them instead of or in conjunction with the SB ELR’s. We can use
them as our BF ELR estimates and calculate BF IBNR with them.
We can then weigh this a priori IBNR against the CL IBNR to
obtain our final “credibility” IBNR estimates. You will find an
interesting discussion of this credibility method in Benktander
[23] and Mack [30].
You can see there are many possibilities, and no single right
method. Any good actuary will want to use as many legiti-
mate methods for which reasonably good information and time
is available, and compare and contrast the estimates from these
methods. As with pricing, it is often informative to see the spread
of estimates derived from different approaches. This helps us un-
derstand better the range and distribution of possibilities, and
may give us some idea of the sensitivity of our answers to
varying assumptions and varying estimation methodologies. A
distribution-free method for calculating the variance of chain-
ladder loss reserve estimates is described in Mack [29].
Alternative Estimation Methodologies

Suppose, as above, we have only the summary development
triangles of reported or paid claims. We can obtain maximum
likelihood estimates of the parameters for a continuous lag model

(such as the Gamma used above) by treating the increments for
each development interval as grouped data, exactly as discussed
by Hogg and Klugman [11] for claim severity. The reported
claim dollars for each accident year can be considered to be a
sample from a truncated model (unknown tail). A slight practical
problem here may be negative increments. But for the estima-
tion, the time intervals for the grouping of the data need not
necessarily all be one-year periods, so the intervals can always
be adjusted to avoid negative increments.
Various statistical and reasonableness tests can then help us
decide which lag model best describes the data, and which we
believe will best predict future claims development. This model
with the fitted parameters can then be used to estimate the prob-
ability distribution for the IBNR, calculate various statistics such
as the expectation and standard deviation, and predict IBNR claim
emergence.
More advanced models, as discussed in the chapter on loss
reserving, could use claim counts and amounts and construct
a stochastic model for the whole claims development process.
These models are useful for very long-tail exposures. They can
give us more information and provide insight we can’t get from
the simpler models discussed above. But they are very much
more complicated, and look to many actuaries like black boxes.
A problem with increasingly sophisticated methodologies is that
the answers may become less intuitive and may be much more
difficult for the actuary to understand and explain to manage-
ment.
An advantage of using a claim count/claim severity model is
that we can contemplate intuitively satisfying models for various
lag distributions, such as the time from loss event occurrence
until first report and the lag from report until settlement. We can
then connect these lags with appropriate models for the dollar
reserving and payments on individual claims up through set-
tlement. For surveys of some advanced methodologies, see the
chapter on Loss Reserving and various Advanced Techniques

sessions in Casualty Loss Reserve Seminar transcripts [26].
Monitoring and Testing Predictions

A loss reserve or an IBNR reserve is derived from a hypoth-
esis about future claims settlements for past events. In order to
validate our methodology, we must test our predictions against
actual future experience. Monitoring and testing quarterly claims
run-off against predictions may provide early warning of prob-
lems.
For short-tailed and medium-tailed lines, this can be fairly
simple. As long as current accident year claims can be reason-
ably separated from past accident year run-off, the run-off can
be compared with the previous year-end reported open and IBNR
reserves. For long-tailed lines, slightly more sophisticated com-
parisons are necessary.
Perhaps the best way to describe a simple IBNR monitoring
methodology is through the use of an example. We will consider
GL exposure for treaty casualty excess contracts with attachment
points in RAA range 4. RAA range 4 currently has attachment
points between $1,251,000 to $3,500,000. We, of course, deflate
this range back via estimated GL ground-up claims severity trend,
so that we include the appropriate contracts from earlier years in
our database.
Table 7.68 is one possible format. Columns 4 and 5 are a
Gamma distribution with mean 5 years and standard deviation 3
years. Column 5 is column 4 adjusted by one half year.
Interpreting Table 7.68 is difficult. Column 8 tells us there was
$809,000 more claims emergence in the first half of 2001 than
the $2,437,000 = $63,240,000 ! $60,803,000 expected. This is
about 33% more than expected. Is this purely random? Or does
it indicate that the beginning IBNR reserve was too small, or the
lags too short? We will want to watch these claims emergence
tests each quarter to see if a positive difference persists. This

might lead us to decide to lengthen the lags for this exposure.
Reinsurance Loss Reserving: Final Comments

We have seen some examples of how standard actuarial meth-
ods and some not-so-standard actuarial methods apply to reinsur-
ance loss reserving. We must remember that there is no one right
way to estimate reinsurance loss reserves. But there are many
wrong ways. Common actuarial methods should be used only to
the extent they make sense. To avoid major blunders, the actuary
must always understand as well as possible the types of rein-
surance exposure in the reinsurance company’s portfolio. The
differences from primary company loss reserving mainly involve
much less specificity of information, longer report and settlement
timing delays, and often much smaller claim frequency together
with much larger severity, all inducing a distinctly higher risk sit-
uation. But with this goes a glorious opportunity for actuaries to
use fully their theoretical mathematical and stochastic modeling
abilities and their statistical data analytical abilities.
REFERENCES 465
REFERENCES
Included here are also works in addition to those cited in the

text.
References for Introduction

[1] Cass, R. Michael, Peter R. Kensicki, Gary S. Patrik,
and Robert C. Reinarz, Reinsurance Practices (2nd ed.),
Malvern, PA: Insurance Institute of America, 1997.
[2] Elliott, Michael W., Bernard L. Webb, Howard N. Ander-
son, and Peter R. Kensicki, Principles of Reinsurance (2nd
ed.), Malvern, PA: Insurance Institute of America, 1997.
[3] Reinarz, Robert C., Janice O. Schloss, Gary S. Patrik, and
Peter R. Kensicki, Reinsurance Practices, Malvern, PA: In-
surance Institute of America, 1990.
[4] Strain, Robert W., Reinsurance, New York: The College of
Insurance, 1980.
[5] Webb, Bernard L., Howard N. Anderson, John A. Cook-
man, and Peter R. Kensicki, Principles of Reinsurance,
Malvern, PA: Insurance Institute of America, 1990.
[6] Webb, Bernard L., Connor M. Harrison, and James J.
Markham, Insurance Company Operations, Vol. II, Chap-
ters 8 and 9, Malvern, PA: American Institute for Property
and Liability Underwriters, 1992.
References for Reinsurance Pricing

[7] Bear, Robert A., and Kenneth J. Nemlick, “Pricing the Im-
pact of Adjustable Features and Loss Sharing Provisions of
Reinsurance Treaties,” Proceedings of the Casualty Actuar-
ial Society, 1990, Vol. 77.
[8] Daykin, C. D., T. Pentikäinen, and M. Pesonen, Practi-
cal Risk theory for Actuaries, London: Chapman and Hall,
1994.
[9] Bühlmann, Hans, Mathematical Methods in Risk Theory,
New York: Springer-Verlag, 1970.
[10] Heckman, Philip E., and Glenn G. Meyers, “The Calcula-

tion of Aggregate Loss Distributions From Claim Severity
and Claim Count Distributions,” Proceedings of the Casu-
alty Actuarial Society, 1983, 70:22–61; discussion by Gary
Venter, 1983, 62–73; exhibits by Heckman and Meyers,
1983, 71:49–66.
[11] Hogg, Robert V., and Stuart A. Klugman, Loss Distribu-
tions, New York: Wiley, 1984.
[12] Klugman, Stuart A., Harry H. Panjer, and Gordon E. Will-
mot, Loss Models: From Data to Decisions, New York: Wi-
ley, 1998.
[13] Ludwig, Stephen, J., “An Exposure Rating Approach to
Pricing Property Excess-of-Loss Reinsurance,” Proceed-
ings of the Casualty Actuarial Society, 1991, 78:110–145;
discussion by Sholom Feldblum, 1993, 80:380–395.
[14] Mashitz, Isaac, and Gary S. Patrik, “Credibility for Treaty
Reinsurance Excess Pricing,” CAS Discussion Paper Pro-
gram, 1990, 317–368.
[15] Miccolis, Robert S., “On the Theory of Increased Limits
and Excess of Loss Pricing,” Proceedings of the Casualty
Actuarial Society, 1977, 64:27–59; discussion by Sheldon
Rosenberg, 1977, 60–73.
[16] Panjer, Harry H., “Recursive Evaluation of a Family of
Compound Distributions,” ASTIN Bulletin, 1981, 12, 1:22–
26.
[17] Papush, Dmitry, Gary S. Patrik, and Feliks Podgaits, “Ap-
proximations of the Aggregate Loss Distribution,” CAS Fo-
rum, Winter 2001, 175–186.
[18] Patrik, Gary S., “Estimating Casualty Insurance Loss
Amount Distributions,” Proceedings of the Casualty Actu-
arial Society, 1980, 67:57–109.
[19] Patrik, Gary S., and Russell T. John, “Pricing Excess-of-
Loss Casualty Working Cover Reinsurance Treaties,” CAS
Discussion Paper Program, 1980, 399–474.
REFERENCES 467
[20] Stanard, James N., and Russell T. John, “Evaluating the

Effect of Reinsurance Contract Terms,” Proceedings of the
Casualty Actuarial Society, 1990, 77:1–41.
[21] Steeneck, Lee R., “Loss Portfolios: Financial Reinsur-
ance,” Proceedings of the Casualty Actuarial Society, 1985,
72:154–167.
[22] Venter, Gary G., “Transformed Beta and Gamma Distribu-
tions and Aggregate Losses,” Proceedings of the Casualty
Actuarial Society, 1983, 70:156–193.
References for Reinsurance Loss Reserving

[23] Benktander, Gunnar, “An Approach to Credibility in Cal-
culating IBNR for Casualty Excess Reinsurance,” The Ac-
tuarial Review, 1976, CAS.
[24] Bühlmann, Hans, Rene Schnieper, and Erwin Straub,
“Claims Reserves in Casualty Insurance Based Upon a
Probabilistic Model,” Bulletin Association of Swiss Actuar-
ies, 1980.
[25] Bornheutter, Ronald L., and Ronald E. Ferguson, “The Ac-
tuary and IBNR,” Proceedings of the Casualty Actuarial So-
ciety, 1972, 59:181–195.
[26] Casualty Loss Reserve Seminar Transcript, CAS, various
years, or the CAS Web Site, casact.org.
[27] Hachemeister, Charles A., “A Stochastic Model for Loss
Reserving,” International Congress of Actuaries, 1980.
[28] John, Russell T., “Report Lag Distributions and IBNR,”
Casualty Loss Reserve Seminar Transcript, 1982.
[29] Mack, Thomas, “Distribution-free Calculation of the Stan-
dard Error of Chain Ladder Reserve Estimates,” ASTIN Bul-
letin, 1993, 23, 2:213–225.
[30] Mack, Thomas, “Credible Claims Reserves: The Benk-
tander Method,” ASTIN Bulletin, 2000, 30, 2:333–347.
[31] Patrik, Gary S., “An Actuarial Procedure for Estimating a
Reinsurance Company’s IBNR,” IASA, 1978.
[32] Patrik, Gary S., “Loss Reserving Problems for New or

Small Reinsurers,” Casualty Loss Reserve Seminar Tran-
script, 1982.
[33] Pinto, Emanuel, and Daniel F. Gogol, “An Analysis of Ex-
cess Loss Development,” Proceedings of the Casualty Actu-
arial Society, 1987, 74:227–255.
[34] Reinsurance Association of America, Historical Loss De-
velopment Study, published bi-annually (odd years).
[35] Weissner, Edward W., “Estimation of Report Lags by the
Method of Maximum Likelihood,” Proceedings of the Ca-
sualty Actuarial Society, 1978, 65:1–9.
[36] Weissner, Edward W., “Evaluation of IBNR on a Low
Frequency Book Where The Report Development Pattern
Is Still Incomplete,” Casualty Loss Reserve Seminar Tran-
script, 1981.
APPENDIX 469
APPENDIX A
PARETO DISTRIBUTION
1. Support: X > 0
2. Parameters: b > 0, q > 0
3. C.d.f.:
' (q
b
F(x - b, q) = 1 !
b+x
4. P.d.f.:
qb q
f(x - b, q) =
(b + x)q+1
5. Moments:
k b k (k!)
E[X - b, q] = for k < q
(q ! 1)(q ! 2)(q ! k)
6. b
E[X - b, q] =
q!1
7. qb 2
Var[X - b, q] =
(q ! 2)(q ! 1)2
8. Censored (from above) c.d.f.: (general definition)

&
F(x) for x < c
F(X; c) =
1 otherwise
9. Censored moments: If q ! k is not an integer, then

' (q
k bk (k!) b
E[X ; c - b, q] = !q
(q ! 1)(q ! 2) : : : (q ! k) (b + c)
& ' (
(b + c)k k!
" + * * * + (!1)i
q!k (i!)(n ! i)!
, -
bi (b + c)k!i
" + ***
q!k+i
, - , -)
bk ck
+ (!1)k !
q q
10. Censored expectation:

& ' (q!1 )
b
E[X; c - b, q] = E[X] 1 ! for q > 1
b+c
11. Conditional probability:

' (q
b+x
Prob[X > y - X > x] =
b+y
12. Truncated (from below-conditional) distribution:

Definition: X(d) = X ! d for X > d
Then X(d) is Pareto with parameters b + d, q:
FX(d) (x) = Prob[X(d) & x]
= 1 ! Prob[X(d) > x]
= 1 ! Prob[X > x + d - X > d]
' q (
b+d
=1!
b + (x + d)
' (q
b+d
=1!
(b + d) + x
= F(x - b + d, q)
APPENDIX 471
13. Trended distribution:

Definition: Y = tX t>0
Then Y is Pareto with parameters tb, q:
FY (y) = Prob[Y & y]
* +
y
= Prob X &
t
. 1q
/ b 2
= 1!/
0 ' (2
y 3
b+
t
' (q
tb
=1!
tb + y
= F(y - tb, q)
APPENDIX B
CATASTROPHE LOSS MODEL: EXAMPLE 7.24
For simplicity, the catastrophe distribution for Example 7.24,

property quota share pricing, is modeled with the following as-
sumptions.
1. N = cat. event count + Poisson[3]
So,
E[N] = 3 = Var[N]
2. X = cat. event severity (as loss ratio) + Pareto[8%, 3]

(see Appendix A)
So,
8%
E[X] = = 4%
3!1
2 2(0:08)2
E[X ] = = 0:0064
(3 ! 1)(3 ! 2)
3. L = annual aggregate cat. loss

Using the standard risk theoretic model described in Ap-
pendix G, we have:
E[L] = E[N]E[X]
= 3(4%) = 12%
Var[L] = E[N]E[X 2 ]
= 3(0:0064) = 0:0192
SD[L] = (0:0192)1=2 = 13:86%
APPENDIX 473
APPENDIX C
GAMMA DISTRIBUTION
1. Support: X > 0
2. Parameters: ® > 0, ¯ > 0
3. P.d.f.:
' (
® !1 ®!1 !x
f(x - ®, ¯) = [¯ ¡ (®)] x exp
¯
4 )
where ¡ (®) = x®!1 e!x dx = (® ! 1)¡ (® ! 1)
0
4. E[X - ®, ¯] = ®¯
5. Var[X - ®, ¯] = ®¯ 2
6. Mode = ¯(® ! 1) if ® # 1
7. If ® = 1, then Exponential distribution.
8. The Gamma distribution is used in the text to model
the distribution of aggregate claims. Recent research (Pa-
push, Patrik and Podgaits [17]) has shown that the sim-
ple 2-parameter Gamma distribution is a more accurate
representation of the aggregate claims distribution in-
duced by the standard risk theoretic model (Appendix
G) than is the more commonly used Lognormal distri-
bution. Also, see Venter [22].
APPENDIX D
ELCF CALCULATION: TABLE 7.34
For simplicity, the claim indemnity severity model underlying

Table 7.34 is assumed to be Pareto, with parameters b = 5,000
and q = 1:1. Then Formula 7.33 is used to adjust for ALAE added
to indemnity loss, with ° = 20%, and for a clash loading, ± =
5%. The following table shows the calculation of the increased
limits factors using Appendix A, #10 to compute the censored
expectation at each limit.
Pareto[5,000, 1.1] Increased Limits Factors

(1) (2) (3)
ILF w/o risk load
Pareto and w/o ALAE
Policy Limit E[X; PLim - 5,000, 1:1] (2) ( (2[$100,000])
$100,000 $13,124 1.0000
250,000 16,255 1.2386
300,000 16,854 1.2842
500,000 18,484 1.4084
833,333 20,041 1.5271
1,000,000 or 20,579 1.5681
more
APPENDIX 475
APPENDIX E
EXCESS EXPECTED CLAIM SIZE: TABLE 7.36

For simplicity, the claim indemnity severity model underly-
ing Table 7.36 is the Table 7.34 Pareto adjusted to add ALAE
to indemnity by altering the Pareto b parameter by the ° = 20%
ALAE loading (new b = 1:2 " 5,000 = 6, 000) and leaving q =
1:1. Since the indemnity loss is censored by policy limits, the
20% add-on for ALAE is simply modeled by adjusting the com-
bined indemnity-plus-ALAE limits upward by 20%. Using Ap-
pendix A, #12 (truncated below at $300,000) and #13 (adding
20% ALAE is like trending by 20%), we get the following result.
1. Expected excess claim severity, over attachment point d
and subject to reinsurance limit RLim, for policy limit ¸:
(1 + ±)E[X(d); ¸(1 + °) ! d - (1 + °)b + d, q]
for d(1 + °)!1 & ¸ & (RLim + d)(1 + °)!1
= (1:05)E[X; (1:2)¸ ! 300,000 - 306,000, 1:1]
for 250,000 & ¸ & 833,333
= (1:05)(306,000=0:1)
& ' (0:1 )
306,000
" 1!
6:000 + (1:2)¸
& ' (0:1 )
306,000
= (3,213,000) 1 !
6:000 + (1:2)¸
2. If ¸ = 300, 000, we computed

& ' (0:1 )
306,000
= (3,213,000) 1 !
366,000
= (3,213,000)(1 ! 0:9823)
= (3,213,000)(0:0177)
= 57,016
APPENDIX F
CREDIBILITY CONSIDERATIONS FOR EXPOSURE RATING AND

EXPERIENCE RATING
We further discuss the items in Tables 7.41 and 7.42.

TABLE 7.41
Exposure Loss Cost Estimate
' The accuracy of the estimate of RCF, the primary rate correction
factor, and thus the accuracy of the primary expected loss cost
or loss ratio
The accuracy of the estimate of RCF for the rating year, for
each line of business, determines the accuracy of the estimate of
the primary expected loss cost or loss ratio, RCF " PCPLR. And
this estimate is critical because the exposure rating estimate of
the excess loss cost for each line of business is proportional to
the loss ratio estimate.
' The accuracy of the predicted distribution of subject premium
by line of business
If this prediction of the mix of business is accurate, and the
estimates of loss cost for each line are accurate, then the esti-
mate of the overall loss cost should be accurate. If the cedant
has a fairly stable mix of business, and no plans for substantive
changes, a fairly accurate prediction can be made of the mix of
business for the rating year.
' For excess coverage, the accuracy of the predicted distribution of
subject premium by increased limits table for liability, by state
for workers compensation, or by type of insured for property,
within a line of business
Within many liability lines, there are sublines with different
increased limits tables, and thus different increased limits factors.
APPENDIX 477
When rating excess coverage, it is clear that the more accurate

our estimate of the distribution of primary premium volume by
table, the more accurate our exposure rating estimate of loss cost.
Since many primary companies do not keep accurate statistics on
these distributions, we often use industry statistics, which may
not be very accurate for the particular company.
Likewise, for workers compensation, there are large differ-
ences in excess exposure by state.
For property, clearly the type of insured (occupancy, construc-
tion, protection) are also important in determining excess expo-
sure.
' For excess coverage, the accuracy of the predicted distribution of
subject premium by policy limit within increased limits table for
liability, by hazard group for workers compensation, by amount
insured for property
For excess coverage, the policy limit, hazard group or amount
insured (or MPL or PML) clearly affect the amount of expo-
sure to excess claims. Thus, the more accurate these estimates
are, the more accurate the overall estimate of excess exposure to
loss. Again, since many primary companies do not keep accurate
statistics on these distributions, we often use industry statistics,
which may not be very accurate for the particular company. If
there is a substantial change in the limits written during the rating
year, the accuracy of the estimated excess loss cost is reduced.
' For excess coverage, the accuracy of the excess loss cost factors
for coverage above the attachment point
This is obviously important. If your excess loss cost factors
are based upon ISO’s increased limits factors or claims severity
data, you must recognize that, since ISO’s increased limits data
are mainly from policies with limits at or below $1,000,000, their
increased limits factors may not be very accurate for limits above
$1,000,000. Therefore, the higher the attachment point, the less
accurate the exposure rating estimate of loss cost. Estimates of
property or casualty claims severity above $1,000,000 are subject

to great uncertainty.
' For excess coverage, the degree of potential exposure not con-
templated by the excess loss cost factors
Neither the excess exposure arising from the clash of separate
policies or coverages, nor from stacking of limits, are contem-
plated by bureau increased limits factors. Thus, some adjustment
must be made, by using a clash loading factor, to price these ad-
ditional exposures. Obviously, a great deal of judgment is used.
Another part of this problem is the pricing of excess exposure
for which there are no bureau increased limits factors, such as
umbrella liability, farm owner’s liability, or various professional
liability lines.
For property coverage, most excess loss cost factors are de-
rived from fire claims severity statistics. There is then the as-
sumption that roughly the same claim severity applies to other
perils.
If these other exposures are known to be a minor part of the
overall exposure, the same loss cost rates estimated for simi-
lar exposures may be used to estimate the loss costs for these
minor exposures. This is obviously subject to judgment and un-
certainty.
TABLE 7.42
Experience Loss Cost Estimate
' The accuracy of the estimates of claims cost inflation

This has an obvious effect upon the estimates of ground-up
or excess loss cost. Claims cost inflation is usually estimated
from broad insurance industry data, because most companies and
many individual lines don’t have enough claims data in order to
make accurate estimates. Historical inflation trends should be
APPENDIX 479
modified for anticipated economic and societal changes. There

is a lot of judgment involved in these estimates and in the use
for particular companies and lines.
' The accuracy of the estimates of loss development
Historical claims development statistics for a particular line or

company are often chaotic, especially for excess coverage. Thus,
we often select claims development lags based upon broader
data. These lags may not accurately reflect the development po-
tential of the claims for the exposure being evaluated. And all
historical data may not accurately reflect the future claims de-
velopment potential.
' The accuracy of the subject premium on-level factors
As with the estimate of RCF for exposure rating, the accuracy

of the estimates of the premium on-level factors has a direct effect
upon the accuracy of the experience rating estimate of the loss
cost. Although most companies have records of their manual
rate changes, most do not keep good statistics on their rating
plan deviations from manual rates. Therefore, broader insurance
data must be used. Also, further rate deviations occur during soft
markets when, for competitive reasons, some exposures may be
undercounted or more optimistic rate credits are given to retain
business.
' The stability of the loss cost, or loss cost rate, over time
It should be obvious that the greater the fluctuation in the his-

torical loss cost or loss cost rate from year to year, the greater
the sample error in the estimate of the mean, and also the
greater the difficulty in measuring any possible trend in the data.
This is especially the case for excess experience rating, where
there may be few historical claims, so that the so-called pro-
cess risk generates a high coefficient of variation (see Appendix
G).
' The possibility of changes in the underlying exposure over time

Experience rating depends upon our ability to adjust past
claims and exposure data to future cost level. If there have been
significant changes in the book of business of the cedant, such as
writing a new line, the experience rate will reflect these changes.
Thus it must be adjusted, and this is, of course, subject to error.
The less certain you are about the stability of the mix of business
over time, the less certain you should be about the experience
rate.
' For excess coverage, the possibility of changes in the distribution
of policy limits over time
It is usually true that policy limits increase over time to keep
pace with the inflation in the values being insured, thus with the
increase in claim severity. To the extent this is not true, the claim
severity data must be adjusted for the slower or faster change in
limits.
APPENDIX 481
APPENDIX G
THE STANDARD RISK THEORETIC MODEL
We further discuss Formulas 7.43 and 7.44.
1. Aggregate loss: L = X1 + X2 + * * * + XN
where
L = rv (random variable) for aggregate loss

N = rv for number of claims (occurrences, events)
Xi = rv for the dollar size of the ith claim
This formula is true no matter what form the probability
models have.
2. Assumptions:
2.1. The Xi are independent and identically distributed,
with c.d.f. FX (x) = Prob[X & x]
2.2. N is independent of the Xi ’s, and has p.d.f. pN (n) =
Prob[N = n].
3. Then, the c.d.f. for L has the following form:

FL (x) = §N pN (n)FX,n (x)
where FX,n (x) = Prob[X1 + X2 + * * * + Xn & x]
convolution
4. Thus, each noncentral moment of L can be written:

4
k
E[L ] = xk dFL (x)
4
= xk d$§N pN (n)FX,n (x)%
4
= §N $pN (n) xk dFX,n (x)%
= §N $pN (n)E[(X1 + X2 + * * * + Xn )k ]%
5. So, in particular:
E[L] = §N pN (n)E[X1 + X2 + * * * + Xn ]
= §N pN (n)$nE[X]%
= E[X]$§N pN (n)n%
= E[N]E[X]
E[L2 ] = §N pN (n)$E[(X1 + X2 + * * * + Xn )2 ]%
= §N pN (n)$E[§i,j (Xi Xj )]%
= §N pN (n)$§i,j (E[Xi Xj ])%
= §N pN (n)$§i E[Xi2 ] + §i.=,j (E[Xi Xj ])%
= §N pN (n)$nE[X 2 ] + §i.=,j (E[Xi ]E[Xj ])%
= §N pN (n)$nE[X 2 ] + (n ! 1)nE[X]2 %
= $§N pN (n)n%E[X 2 ] + $§N pN (n)n2 %E[X]2
! $§N pN (n)n%E[X]2
= E[N]Var[X] + E[N 2 ]E[X]2
6. Thus, Var[L] = E[N]E[X 2 ] + (Var[N] ! E[N])E[X]2
7. With patience, you can also show that:
E[(L ! E[L])3 ] = E[N]E[(X ! E[X])3 ]

+ E[(N ! E[N])3 ]E[X]
+ 3 Var[N]E[X]Var[X]
APPENDIX 483
8. If N is Poisson, then:
Var[L] = E[N]E[X 2 ]
E[(L ! E[L])3 ] = E[N]E[X 3 ]
APPENDIX H
ABBREVIATIONS
' ACR additional case reserve
' ALAE allocated loss adjustment expense
' BF Method Bornheutter–Ferguson method of loss
development
' CL Method chainladder method of loss development
' ELCF excess loss cost factor
' ELR expected loss ratio
' E[X;c] expected loss cost up to censor c
' IBNER incurred by not enough reserve
' L random variable for aggregate loss
' MPL maximum possible loss
' N random variable for number of claims
' PML probable maximum loss
' PCP primary company premium
' PCPLR primary company permissible loss ratio
' PV present value
' RBF reinsurance brokerage fee
' RCF rate correction factor
' RCR reinsurance ceding commission rate
' RDF reinsurer’s loss discount factor
' REP reinsurer earned premium
' RELC reinsurance expected loss cost
' RIXL reinsurer’s internal expense loading
' RLim reinsurance limit
' RP reinsurance premium
' SB Method Stanard–Bühlmann (Cape Cod) method of
loss development
' TER target economic return
' X random variable for size of claim
' X(d) random variable X excess of truncation
point d

Reinsurance

Uploaded by

Copyright:

Available Formats

Reinsurance

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Reinsurance

Uploaded by

Copyright:

Available Formats

Chapter 7

insurance system. Just as an individual or company purchases an

The Functions of Reinsurance

Financial Results Management

reinsurer to the cedant until the cedant’s surplus is large enough

The Forms of Reinsurance

Property certificate coverage is sometimes written on a pro-

Facultative Automatic Agreements or Programs

occurring during the treaty term or those occurring on policies

Treaty Proportional Covers

A surplus-share treaty also reinsures a fixed percentage of

Treaty Excess Covers

A higher exposed layer excess treaty attaches above the work-

1. extra-contractual-obligations (i.e., bad faith claims)

Aggregate Excess, or Stop Loss Covers

catastrophes are often excluded, or there may be a per-occurrence

Finite, or Nontraditional, Reinsurance Covers

discussion has sufficient risk transfer to pass FAS 113 require-

The first typical form for finite reinsurance is a financial pro-

A loss portfolio transfer is also a very prevalent form for finite

A funded aggregate excess cover is, as you might expect, an

management, it may provide strong stabilization of the cedant’s

A Typical Reinsurance Program

The Cost of Reinsurance to the Cedant

Line of Business Type of Reinsurance

Fire and Allied Lines 1. Proportional and excess facultative certificates to

tude of capacity and financial support, but it is practically influ-

ance premium and is paid by the reinsurer. Offsetting this cost

Lost Investment Income

Additional Cedant Expenses

Balancing Costs and Benefits

Reinsurance Introduction: Final Comments

definition of “loss,” commission arrangements, premium and loss

Reinsurance Pricing Methods

will discuss a few actuarially sound methods. In general, the

A Flat Rate Reinsurance Pricing Formula

return in the simple case where there are no later commission

Formula 7.2: A Flat Rate Reinsurance Pricing Formula

Let’s break down this somewhat complicated-looking for-

underwriting and claims handling effort, and thus inter-

PVRELC = $100,000 (calculated by actuarial analysis

Later we will see cases where the commission or premium is

Thirteen-step Program to Reinsurance Pricing Happiness

1. Gather and reconcile primary exposure, expense and

9. Estimate an experience expected loss cost, PVRELC,

the ceding commission fairly covers the cedant’s expenses, but

The Actuary’s Role

1. Being sure that the facultative underwriters are provided

uation of the accumulation of catastrophe exposure, and also in

Formula 7.6: Reinsurance Expected Loss Cost

RCF = rate correction factor

Suppose that the parameters are given by b = 0:1 and q = 2.

Formula 7.10: Expected Claim Severity (as a fraction of MPL)

Formula 7.11: Expected Claim Count (ground-up)

Formula 7.12: Expected Claim Severity in the Layer

Formula 7.13: Expected Claim Severity in the Layer

3 The factors in a Lloyd’s-type table, mentioned in a previous footnote, would technically

Then we can calculate the technical reinsurance pure premium

To squeeze out more information about the reinsurance cov-

Formula 7.16: Expected Reinsurance Claim Count