OECD Task Force on the
Measurement of Non-life
Insurance Production in the
context of Catastrophes
Final report
F. Lequiller, moderator
Insurance Task Force: Final Report



This is a session for decision
Experts are requested to give their
opinion of the recommendations of
the report, for further forwarding to
the ISWGNA
And submission to the February 2004
session of the AEG, for inclusion in
the next edition of the SNA
Introduction



The current recommendation of the SNA
leads to uninterpretable movements in the
current price value of non life insurance
output and consumption, in particular in
the context of catastrophes
In general, NA compilers neutralise this
unwelcome impact on volume measures.
However, the impact is as unwelcome on
price indices of output or consumption.
Introduction



In the USA, in the third quarter of 2001 (9/11
terrorist attack), the implementation of the
current SNA recommendation led to a decrease
to the price index of household consumption of
0.3%, and an increase of 0.2% of the price of
GDP: ???
In France, the implementation of the current
recommendation on the fourth quarter on 1999
(exceptional storm) would have decreased total
household consumption by 1.3%: ????
The objective of the task force is to avoid such
uninterpretable results
General principle




In the SNA the output of insurance is to « provide
financial protection against the occurence of
specified events »,
However, the SNA recognises that there is no
explicit charge for this output,
Therefore one must use an indirect estimate
based on the formula (1): actual premiums
earned plus premium supplements minus
claims due.
The task force confirmed that, conceptually, the
SNA should not be changed. In particular it
rejected other conceptually based measures of
output (such as premiums).
General principle (continued..)




However, the task force rejected the
mechanical implementation of formula (1)
The measure of the concept of insurance
output should not be affected by the
occurence of the risk.
The unusual difference between premium
and claims in a given period is not to be
attributed to a movement of output in this
period.
The volatility of claims should not affect
the measure of production
Recommendation 1
The concept of insurance service in
the SNA should be preserved.
However, it should be made clearer
that the formula proposed in the SNA
is only the basis for an indirect
estimate of the value of the insurance
service. The measure of the
production of insurance services
should not be affected by the
volatility of claims. Conceptually,
neither the volume nor the price of
insurance services is affected by the
volatility of claims.
The principle of using expectations




Insurers fix their premiums based on their
expectations of the occurence of risks, and of
their other revenues (of which premium
supplements).
Based on these expected claims and expected
premium supplements, the insurer is capable of
fixing the level of premium that will normally
cover its administrative costs and its normal
profit.
The task force was therefore convinced that the
measurement of the SNA service margin
(premiums plus supplements minus claims)
should be based on such expectations.
This is the best way to obtain a measure of
output and consumption consistent with the
concept aimed by the SNA.
The principle of using expectations

Recommendation 2: expected claims and

Formula (2): [Actual premiums earned [i.e.
expected premium supplements should replace
actual claims and actual investment income in the
calculation of the output and consumption of non-life
insurance services. This applies for all claims –
regular, catastrophic and unexpected – and for all
years. This applies for all investment income.
Formula (1) should be adapted to this
recommendation and become:
premiums receivable less changes in the reserves
due to pre-payment of premiums] + Expected
premiums supplements - Expected claims due.
Estimation of expected claims



If insurers were willing to transmit
the amount of their (confidential)
expected claims, it could be used by
national accounts compilers
However, a macro estimate is more
realistic
Several methods were discussed to
estimate expected claims
Estimation of expected claims:
the statistical method







The objective was to find a good and practical way of
estimating the mathematical expectation of claims
The most natural way is to use past claims (including the year
under review) to forecast future claims
A practical statistical method is to use smoothed past claims,
The use of a non centered moving average was recommended,
so that the method does not by itself lead into systematic
revisions
The use of geometric weighted moving average is a workable
solution
In the USA, the BEA implemented a method based on
smoothing loss ratios rather than smoothing directly claims.
Expected claims are obtained by applying the smoothed loss
ratio to actual premiums. This method gave better results
regarding the quality of the prediction than by smoothing
directly claims.
However, some participants of the task force found drawbacks
in the use of smoothed loss ratios, and prefer smoothing
directly claims.
Estimation of expected claims:
the statistical method




These smoothing methods can cope with the
« routine » irregularity of claims
However, they cannot cope with catastrophe
claims
In the case of catastrophe claims, these claims
should be (1) estimated, (2) taken out of the
data, (3) split over a very long period (20 years)
(4) re-included manuallly in the statistical time
series of expected claims (ahead of the current
year), taking account of inflation.
Overall, the statistical method includes therefore
two steps: (1) the exclusion of major exceptional
claims (which will be reincluded at the end), (2)
the smoothing of past claims or loss ratios.
Estimation of expected claims:
the statistical method

Recommendation 3: In the absence of actual
expected data from insurers, the first method for
measuring expected claims is a statistical method
based on the smoothing of past claims either through
the calculation of an expected loss ratio derived from
smoothed past loss ratios, and applicable to actual
premiums, or the direct smoothing of past claims.
The smoothing method must include a prior step
which excludes major exceptional claims. These
exceptional claims should be reintroduced ahead of
the current year by splitting them evenly over a very
long period of time, taking account of inflation. The
task force recommended that the method should not
lead to built-in revisions of the data. It should use
past data available at the moment of the theoretical
decision by insurers of the level of their premiums.
This includes the year under study.
Technical provisions



Some delegates in the task force thought
that the use of the full range of
« provisions » included in the insurance
companies’ accounts could help to avoid to
use a statistical method.
The report first discusses these
« provisions ».
It recommends (1) to change the
terminology of the SNA, (2) to extend the
range of provisions included in formula
(1).
Technical provisions


Recommendation 4: It is proposed to align with the
general accounting principles and to label, in the SNA,
the technical liabilities of insurers as “insurance
technical provisions”.
Recommendation 5: It is proposed to include in the
SNA’s definition of provision for unearned premiums
and provisions for claims outstanding the other
special provisions recorded by the insurance
companies. i.e: equalisation provision, provision for
bonuses and rebates, provision for claims
outstanding, provision for incurred but not (enough)
reported incident
The accounting approach



Some delegates in the task force thought that use of
formula (1) using an extended definition of provisions
would be a possible way of approaching the measure
of output and consumption, without having to use an
indirect statistical estimate.
However, there is no practical experience of this
method for the moment. Also, it is recognised that
equalisation provisions are used sometimes more for
fiscal reasons than for equalisation reasons.
Also, equalisation provisions will be unsufficient to
cover catastrophe claims. Companies have therefore
to use their own-funds. Then the formula should be
extended to own-funds. However, there is no practical
rule to analyse when the formula should be extended
to own-funds.
The accounting approach

Recommendation 5: SNA’s formula (1), which
defines the indirect measure of production from
an accounting perspective, should extend the
scope of the provisions to include equalization
provisions and other special provisions destined
to cover unusual claims. Where it can be proved
that movements in these special provisions
compensate the volatility of claims, an
alternative method to estimate the production of
insurance would be to use this extended formula:
[Premiums earned + premium supplements] –
[claims due + addition to, less withdrawal from,
equalization provisions + addition to, less
withdrawal from own funds, when necessary].
Premium supplements

The discussion on premium supplements
covered three issues:
1 Should expected premium supplements
replace observed premium
supplements?
2 Should premiums supplements include
income from special reserves and own
funds?
3 Should premium supplements include
holding gains?
Expected premium supplements


There are arguments in favor of expected
premium supplements but also against.
Recommendation 6: Expected
premium supplements should replace
actual premium supplements in the
formula used to estimate insurance
production. The method used to
estimate expected premium
supplements should be the best
statistical predictor of premium
supplements.
Special reserves and own funds



Special reserves and own funds bring
revenues that insurers take into account
to fix their premiums.
In consequence, and in accordance with
the task force on financial services, it is
proposed to extend the calculation of
premium supplements to all special
reserves and own funds.
This does not mean that they are to be
considered as assets of the insurance
policy holders.
Special reserves and own funds

Recommendation 7: Investment
income from special reserves and
own funds should be included in
the formula that determines
expected investment income.
However, this would not lead
own funds to be classified as
owned by policy holders.
Holding gains and losses



The issue was a secondary one for this
task force, but an essential one for the
task force on financial services.
The task force on financial services did not
recommend to extend the formula of
FISIM to holding gains and losses.
Recommendation 8: In accordance
with the conclusions of the task force
on financial services holding
gains/losses should not be included
in premium supplements.
Incorporating the new measure of
output in the sector accounts
 The move to a measure of output based on
expected claims rather than observed claims
creates an unbalance in the institutional sector
accounts.
 In the current SNA, there are two identities:
• D44 property income distributed to policy holders is
identically equal to D4 property income earned by the
insurance company
• D72 non life insurance claims is identically equal to D71,
net non life insurance premiums
 Combined with the current measure of production,
this ensures that the balancing items (savings, net
lending/borrowing) are relevant.
 Without the addition of adjustement items, the
change in the measure of production proposed by
the task force would lead to irrelevant balancing
items.
The simplest solution




The simplest solution is to accept to decouple one
of the accounting identities: D72 would no more
be equal to D71;
D72 would remain unchanged, equal to actual
non life insurance claims received by policy
holders.
Only D71 would be changed: rather than be
equal to actual non life insurance claims, it will be
equal to expected non life insurance premiums
plus the difference between actual premium
supplements and expected premium
supplements.
With this simple solution, an implicit (positive or
negative) transfer is made in the current
accounts to policy holders, equal to the difference
between D72 and D71.
A possible sophistication, involving
capital transfers



The implicit transfer to policy holders affect their
disposable income (while saving will be equal to
the saving of the current system)
This was seen by some participants as a
drawback, specially when major catastrophes
occur. It is indeed strange that disposable
income, which is a measure of current income, is
affected by exceptional differences between
expected claims and observed claims.
The exceptional nature of this difference would
point out in the direction of a capital transfer.
A consensual recommendation

Recommendation 9: the introduction of expected
claims and expected premium supplements in the
calculation of production will result in a decoupling
of non-life insurance claims (D72) and the
corresponding imputed net non-life insurance
premiums (D71). D71 will be equal to expected
claims plus the difference between actual premium
supplements and expected premium supplements. In
the case of catastrophes, where the difference
between D71 and D72 may be deemed too important
to affect current disposable income of policy holders,
the difference attributed to the catastrophe can be
treated as a capital transfer, to avoid affecting
disposable income.
Re-insurance



Reinsurance is a major way for direct
insurers to deal with exceptional
claims.
It was therefore obvious that the
treatment of reinsurance was to be
discussed by the task force.
Also, the current SNA needed some
clarification on the treatment of
reinsurance
Reinsurance: main findings





The current SNA « consolidated » treatment of
reinsurance has drawbacks
There are two workable solutions to treat
reinsurance: the full net approach, the full gross
approach.
However none of these two approaches resolve,
by themselves the problem of catastrophic losses
in the measure of production
The only solution is to apply the method of
expected claims to both direct insurers and
reinsurers
In this context, the gross method appears to be
the most practical one
Reinsurance

Recommendation 10: The treatment of reinsurance in the SNA has to be revised. The
consolidated approach for re-insurance
transactions of the current SNA should be
replaced by an approach which treats direct
insurers and re-insurers exactly in the same
way. In particular, the formulae to compile
production in both cases should be identical
in their principles. Both should use
expected claims and expected premium
supplements. The gross presentation,
where the production of re-insurance is
consumed by direct insurers, is preferred to
the net approach, where this production is
netted out.
Additional recommendations


Recommendation 11: The commissions paid by the
reinsurer to the direct insurer and the rebates paid
by the direct insurer to the policyholders should be
classified as a negative component of the premiums
earned. The profit sharing paid by the reinsurer to
the direct insurer and the bonuses paid by the direct
insurer to the policyholders should be classified as a
form of other income transfers.
Recommendation 12: the terminology used in the
SNA should be made closer to the one used in the
insurance industry. In particular, “Net premiums”
should be changed to “Risk premiums” and ¨Claims
due” to ¨Claims incurred”.
Conclusion

Experts from OECD countries are
requested to comment these
recommendations, and in priority:
1 Principles (recommendations 1&2)
2 Estimation of expected claims
(recommendation 3)
3 Integration in the accounting
framework (recommendation 9)