Consultation Paper No. 49 - CEIOPS-CP-49/09, 2
July 2009
Draft CEIOPS
Advice for Level 2 Implementing Measures on Solvency II:
Standard formula SCR - Article 109 c, Life
underwriting risk
3.7 Lapse risk
3.7.1 Explanatory
text
Previous advice
3.101. In its “Further advice to
the European Commission on Pillar 1 issues” of March 2007, CEIOPS
recommended the inclusion of an explicit requirement for lapse risk
under the SCR standard formula.
Lapse risk was understood to
arise from unanticipated (higher or lower) rate of policy lapses,
terminations, changes to paid-up status (cessation of premium
payment) and surrenders.
3.102. In the advice document it
was noted that the complex dependence structure of lapse risk is
difficult to model in a modular approach to the standard
formula.
Lapse risk in QIS4
3.103. The QIS4 approach
to the SCR standard formula included a lapse risk submodule in the
life underwriting risk module.9 The calculation of the capital
requirement for lapse risk was based on three scenarios:
• a
permanent increase of lapse rates by 50%;
• a permanent
decrease of lapse rates by 50%; and
• a mass lapse event
where 30% of the policies are surrendered.
Treatment of lapse
risk in the scenario calculations
3.104. For life insurance,
the calculation of the SCR is essentially scenario-based:
all
life insurance underwriting risks and all market risks except spread
risk and concentration risk are quantified by means of scenario
analysis.
Moreover, the loss-absorbing capacity of technical
provisions is determined by means of a scenario.
3.105. The
definition of these scenarios is usually understood as follows:
assume that a specific change relating to a certain risk (e.g.
change of interest rates or a change of equity prices or a change of
mortality rates) takes place while all parameters relating to other
risks remain unchanged.
This is the approach that was
applied in QIS4.
Discussions during the exercise in some
markets have shown that this simple and straightforward scenario
definition may lead to unrealistic results which do not fully
reflect the risk that the insurers are exposed to.
The
following examples should illustrate the issue.
Three
examples for lapse risk triggered by other risks
Example 1:
Lapse triggered by the reduction of bonus rates
3.106. In
life with-profit business, many risks can be mitigated by cutting
future bonuses.
The QIS4 report shows that the adjustment
for the risk mitigating effect has a significant influence on the
SCR in many countries.
In the market with the highest impact,
about 75% of the BSCR is reduced by the adjustment for the
loss-absorbing capacity of technical provisions on average.
Under the scenario definition outlined above, this may be a
fair reflection of the undertaking’s profit sharing
systems.
3.107. Nevertheless, for certain kinds of business
the outcome appears to be unrealistic because a significant cut of
future bonuses may change the lapse behaviour of
policyholders.
For example, consider a term insurance where
the extra benefit of the policies is used to reduce the premiums.
A significant cut of extra benefits as a reaction to (for
instance) an equity shock would increase the number of lapses
because many policyholders would rather terminate the contract than
pay a significantly higher premium.
The increase of lapses
would consequently increase the technical provisions because the
business is usually profitable and the future profit of the
terminated treaties cannot be taken into account anymore in the
provision.
Hence, the cut of extra benefits would trigger
lapses which would at least partly thwart the mitigating effect of
profit sharing.
3.108. This effect was not taken into account
under the QIS4 scenario approach because it was assumed that lapse
rates are fixed in the equity scenario.
Moreover, this
effect is also not fully covered by the lapse risk module because
firstly, the lapse effect of the reduction in extra benefits can be
higher than the shocks considered in the lapse sub-module and
secondly, there is a high diversification effect between lapse risk
and equity risk in the standard formula.
According to Annex
IV of the Level 1 text, the correlation factor for market risk and
life underwriting risk is 25%.
However, in the example there
is a causal connection between equity risk and lapse
risk.
Example 2: Lump-sum option triggered by the increase of
interest rates
3.109. In deferred annuity insurance, the
policyholder can often choose between a lump sum and a previously
fixed annuity at expiration of the mortality cover.
The
take-up rate for this option is very likely to be interest rate
sensitive.
3.110. If the market interest rates are
significantly lower than the technical rate that was used to
determine the annuity, it is rational for all policyholders to
choose the annuity.
Therefore, in the interest rate
decrease, scenario the number of policyholders which choose the
annuity would go up, thereby increasing the loss of the
insurer.
3.111. On the other hand, if the market interest
rates are significantly higher than the technical rate more
policyholders are likely to choose the lump sum.
Therefore,
in the scenario of an interest rate increase the reduction of
technical provisions due to discounting with higher rates may be
partly countered by the loss of future profit.
Neither of
both effects is currently allowed for in the SCR.
Example 3:
Reduction of insurance cover triggered by the decrease of interest
rates in health insurance
3.112. In health insurance as
described in Article 204 of the Level 1 text (technical basis
similar to that of life insurance), the insurer can adjust the
premiums according to a specific mechanism in order to take into
account a change in risk factors like health expenses, longevity or
interest rates.
For example, under the interest rate decrease
scenario, this mechanism is usually applied to mitigate the stress.
In line with the scenario definition outlined above, all other risk
factors remain unchanged.
3.113. However, in reality it is
very likely that the increase in premium levels would cause a part
of the policyholders to reduce its insurance cover to compensate for
the financial strain.
This would reduce the future profits
and therefore reduce the mitigating effect of the premium
increase.
3.114. Similar effects may exist in relation to
other kinds of insurance with adjustable premiums.
3.115.
Summarising, the three examples differ in both the triggering event
as well as the option that they affect:
Example Triggering
event Option affected
3.116. An insufficient allowance for lapse risk in the
SCR standard formula has also been noted by stakeholders. It was
criticised that “the main risk in life insurance is not taken into
account in the standard formula”, namely “a policyholder run
touched-off by market, credit, or operational risk”.
Lapse
triggered by deterioration of financial position
3.117.
Another important external trigger of policyholder action is the
deterioration of the undertaking’s financial position.
Such
an event, if it becomes apparent, may cause a mass lapse incident,
thereby reinforcing
the decline of the financial
position.
Proposal to allow for lapse risk in the SCR
scenarios
3.118. The examples demonstrate that the QIS4
approach to lapse risk (and other option take-up risks) does not
take into account that the take-up of the option by the policyholder
may be triggered by other risks or the reaction of the insurer to
other risks.
On the contrary, the QIS4 standard formula
makes the assumption that lapse risk is approximately
independent
from other risks.
Disregarding the dependence of option
take-ups and other risks may lead to a significant underestimation
of the 99.5% confidence level of the SCR.
3.119. It seems
that this deficiency cannot be eliminated by an increase of the
correlation factors that are used to aggregate lapse risk and the
other risks.
In order to model a high dependence between
lapse risk and, for example, interest rate risk in the current
modular structure of the SCR, the correlation factor for market and
life underwriting risk must be increased.
But this would
also increase the modelled dependence between market risk and
mortality risk, longevity risk and CAT risk although these risks are
likely to be less dependent.
Hence, an increase of the
correlation factors may partly remedy the deficiency in relation to
lapse risk, but would lead to an unjustified increase of the
dependence between other risks
3.120. Another way to tackle
the problem would be to relax the scenario definition applied in the
SCR calculations. Instead of changing one parameter (e.g. interest
rates) in the scenario and keeping all other parameters fixed, the
scenario could also allow for adverse changes in option take-up
rates.
For instance, the interest rate decrease scenario
(net of profit sharing) could be defined as
follows:
where
QNAV|downwardshock is the change in the net value of asset and
liabilities due to revaluing all interest rate sensitive instruments
using altered term structures.
The revaluation is done under
the condition that the participant is able to vary its assumptions
on future bonus rates in response to the shock.
Moreover,
the revaluation should allow for any relevant adverse changes in
option take-up behaviour of policyholders in this
scenario.
Practicability of the proposal
3.121. The
proposal should not to cause practical problems.
In
principle, the undertaking already needs to make assumptions about
the behaviour of its policyholders in extreme scenarios in order to
calculate the value of options and guarantees for the best estimate
provisions according to Article 78 of the Level 1 text.
Consistent assumptions can be used in the SCR scenarios to
allow for changed policyholder behaviour.
In case an
approximation is used to value the options and guarantees, it should
be possible to derive assumptions about the stressed option-take up
rates which are consistent with the approximations.
Nevertheless, in some cases it may be helpful to give
market-specific actuarial guidance about the policyholder
behaviour in relation to certain products in order to ensure the
practicability and comparability of the SCR
calculations.
3.122. In line with the proportionality
principle, the proposal includes only the relevant changes in option
take-up rates.
Relation to the lapse risk
sub-module
3.123. According to the proposal, lapse risk is
allowed for twice in the SCR standard formula: in the lapse risk
sub-module and in each scenario which has a relevant adverse effect
on the lapse rates. However this does not give rise to a double
counting of lapse risks.
3.124. Conceptually, two elements of
lapse risk can be distinguished:
A. Misestimate of current
lapse rates
A misestimate of the lapse rates which are
appropriate (this year and in future years) according to the current
situation (i.e. the current interest rates, bonus rates
etc.).
B. Change of lapse rates
The change of lapse
rates owing to a change of the current situation (i.e. a change in
interest rates, bonus rates etc.).
The examples given in
paragraphs 3.106 to 3.116 illustrate this risk.
3.125. An
allowance for lapse risk in the SCR scenarios as proposed above
would only cover lapse risk of type B.
Besides, it would not fully
cover this kind of lapse risk because not all possible changes of
the current situation which affect the lapse rates are reflected in
SCR.
For instance, the effect described in the example of
paragraph 3.116 is not covered by the
SCR scenarios.
3.126. Complementary to this approach, the
lapse sub-module of the life underwriting risk module allows for the
residual lapse risk of type B as well as the complete lapse risk of
type A.
Therefore, the proposal does not lead to double
counting of lapse risk.
However, the calibration of the
lapse scenarios in the lapse sub-module, in particular of the mass
lapse event, should take into account the distinction made paragraph
3.124.
Scope of the lapse risk sub-module
3.127.
According to Article 105(3f) of the Level 1 text, the lapse risk
sub-module covers the risk of
“loss, or of adverse change in the
value of insurance liabilities, resulting from changes in the level
or volatility of the rates of policy lapses, terminations, renewals
and surrenders (lapse risk)”
3.128. This description does not
fully clarify which policyholder options are included in the
sub-module. In particular, the scope of the terms “lapse”,
“termination” and “surrender” is not clear.
For example, in
QIS4 the risk relating to changes to paid-up status were included in
the sub-module because it was interpreted as a partial
termination.
The scope of the module should be clarified at
Level 2 in order to remove potential ambiguity in the calculation of
the lapse risk capital requirement.
3.129. There are two
possible antipodal approaches to the scope of the lapse risk
module:
• A narrow definition of the scope: only those
options are included which either fully renew the insurance cover or
which fully terminate the policy and which are defined as surrender
or lapse options in the terms and conditions of the policy.
•
A wide definition of the scope: all options are included where a
take-up or non take-up reduces the insurance cover.
3.130.
There is a range of possible definitions between these antipodal
approaches.
3.131. The narrow definition would not include,
for example, changes to paid-up status.
If the narrow
definition is applied it may be advisable to add further sub-modules
to the life underwriting risk module in order to cover the risk of
other options.
Moreover, with the narrow definition
arbitrage opportunities appear to be unavoidable.
Policyholder options which slightly differ in their effect
or description may be treated differently in the SCR calculation.
For example, a partial termination option which reduces the
insurance cover by 99% is not included in the narrow definition
although the economic effect is comparable to a full termination
which is included.
3.132. On the other hand, a wide
definition could be a clear and simple way to cover all material
option risks consistently within the structure of the standard
formula.
It would not be necessary to add further
sub-modules to the life underwriting modules which cover other
option risks.
Therefore, the wide definition is
preferable.
3.133. Under the wide definition of scope, the
scenarios which define a permanent decrease and increase of
option take-up rates could be defined as
follows:
lapseshockdown = Reduction of x% in the assumed option
take-up rates in all future years for all policies without a
positive
surrender strain.
Affected by the reduction
are options to fully or partly terminate, decrease,
restrict or suspend the insurance cover.
Where an
option allows the full or partial establishment,
renewal, increase, extension or resumption of insurance
cover,
the x% reduction should be applied to the rate that the option is
not taken up.
lapseshockup = Increase of y% in the assumed option
take-up rates in all future years for all policies with a
positive surrender strain.
Affected by the increase are
options to fully or partly terminate, decrease, restrict
or suspend the insurance cover.
Where an option allows the
full or partial establishment, renewal, increase, extension or
resumption of insurance cover, the y%
increase should be applied
to the rate that the option is not taken up.
3.134. The
surrender strain of a policy is defined as the difference between
the amount currently payable on surrender and the best estimate
provision held.
The amount payable on surrender should be
calculated net of any amounts recoverable from policyholders or
agents e.g. net of any surrender charge that may be applied under
the terms of the contract.
Capital requirements for the three
sub-risks (see paragraph 3.135) should be calculated based on a
policy-by-policy comparison of surrender value and best estimate
provision.
In this context, the term “surrender” should refer to
all kind of policy terminations irrespective of their name in
the
terms and conditions of the policy. In particular, the
surrender value may be zero if no compensation is paid on
termination.
Calculation of the capital
requirement
3.135. The calculation of the capital requirement
for lapse risk in QIS4 was based on three scenarios:
• a
permanent increase of lapse rates;
• a permanent decrease of
lapse rates; and
• a mass lapse event.
3.136. The
capital requirement was obtained as the loss of net asset value
under the most adverse of the three scenarios.
This simple
approach has some shortcomings.
For example, an insurer may
be exposed to the risk of an increase in lapse rates on one part of
its portfolio and a decrease of lapse rates in another part of the
portfolio.
Such situations are not covered by the approach.
However, within the natural limitations of the standard
formula the QIS4 approach appears to be an acceptable solution.
The feedback from the QIS4 participants gives no other
indication.
Calibration
Calibration of lapseshockup and
lapseshockdown
3.137. As lapse rates are not frequently used for
reserving under Solvency I, the empirical basis for a calibration
of the permanent shocks lapseshockup and lapseshockdown is poor
for most markets.
3.138. The QIS4 calibration of the shocks was
based on a study of the UK with profit life insurance market in
2003 performed by order of the British FSA.11 The analysis
resulted in estimates for quantiles of permanent lapse rate
decreases as follows:
3.139.
The quantile produced in the study are lower than the
Solvency II
confidence level of 99.5%.
Nevertheless, by extrapolation of
the above values, the QIS4 calibration of -50% can be justified.
The study does not cover the risk of a permanent increase of
lapse rates, however, in absence of better evidence it is
appropriate to assume a symmetrical stresses for both scenarios and
choose +50% for the increase scenario.
3.140. CEIOPS has
looked for further evidence from other markets. An analysis of the
Polish supervisor on the national life insurance market supports the
above calibration assumptions (see Annex C).
The study shows
that the 99.5% quantile of annual lapse rate deviations from a
long-term mean is between 60% and 100% for increases and between
-60% and -90% for decreases.
As these values are based on an
annual deviation they overestimate the shock of a permanent change.
However, the results indicate that the range of the proposed
calibration is appropriate.
3.141. The lapse shocks were
calibrated on small rates. If the rates are much larger, the
calibration may produce excessive results. Moreover, it needs to
be avoided that the shocked rates exceed 100%.
3.142. Therefore,
the shocked take-up rate should be restricted as follows:
3.143.
The
scenario shocks lapseshockup and lapseshockdown cover the risk of a
misestimation or of a permanent change of lapse rates.
By
contrast, the mass lapse event covers the risk of a temporary and
drastic rise of lapse rates.
The likeliness that
policyholders terminate their policies is increased for a limited
span of time.
The cause for this change in policyholder
behaviour can be of an internal or external nature.
An
internal cause could, for example, be the deterioration of the
financial position of the undertaking or any other event that
significantly affects the reputation of the undertaking or the group
it belongs to.
Examples of external events would be changes in
the economic situation or changes in the tax regulations that
directly or indirectly affect the policies of the
undertaking.
An event in the banking sector comparable to the
mass lapse event would be a “bank run”.
3.144. The
calibration of the mass lapse event should account for the scenario
definition as defined above (paragraph 3.104 ff.).
Where the
change in lapse behaviour is triggered by a change in scenario-based
risk like interest rate risk or equity risk, an allowance in the
mass lapse event is not necessary.
The calibration of the
mass lapse event should only cover those changes in behaviour which
are not triggered by these risks.
3.145.
On the other hand,
the calibration of the mass lapse event has to reflect the fact that
mass lapse is a “catastrophe” type event.
Policyholder
behaviour under extreme conditions is difficult to assess as it can
be determined by complex phenomena like herd behaviour and
self reinforcing mechanisms.
Experience from the banking
sector during the current financial crises shows
(for example
Northern Rock bank run in 2007) that policyholder behaviour can pose
a significant risk to financial institutions.
3.146. Under
Solvency I insurance and reinsurance undertakings are less affected
by lapse risk as the technical provisions for a policy must not be
lower than its surrender value.
But under Solvency II it may
happen that the assets of an undertaking do not cover the surrender
values.
Such insurers are highly vulnerable to mass lapse
events, in particular when their situation becomes
public.
3.147. The empirical basis to calibrate the mass
lapse event is poor.
In the absence of better evidence,
CEIOPS proposes to maintain the QIS4 calibration of 30%
of the sum
of positive surrender strains.
3.148. It has been discussed
whether different types of life insurance policies are affected
differently by mass lapse events: products with significant
guarantees like with-profit products may show a higher persistency
than products with low guarantees like many unit-linked
policies.
3.149. On the other hand, for non- retail
business13, the risk of a mass lapse is substantially greater for
the following reasons:
• Institutional investors tend to be
better informed and would be quick to withdraw funds if there was
any question over the solvency of a firm, particularly if they were
aware that the firm did not have sufficient funds to meet all
claims;
• There are generally no surrender
penalties.
3.150. CEIOPS therefore believes that a higher
calibration of the mass lapse stress is appropriate for this
business.
In the absence of other information, CEIOPS
proposes to use the QIS3 calibration of 70% of the sum of positive
surrender strains.
3.151. At this stage, taking into account
a simple valuation of the mass lapse event, CEIOPS is considering
whether to differentiate further between different insurance
products for the purposes of the mass lapse stress.
3.152.
Question to stakeholders:
a. To the extent that lapse risk
depends on product characteristics, where should the balance between
simplicity of the calculation of the requirement and the reflection
of these characteristics in the
requirement be?
What are
stakeholder views on the product characteristics impacting on mass
lapse experience and how could these characteristics be defined for
the purposes of the mass lapse stress?
b. If stakeholders
would like to propose an alternative calibration, it would be
helpful if a justification (in particular a description of the
empirical and prospective basis for the calibration) could be
provided.
Simplifications
Calculation on
policy-by-policy basis
3.153. In case the best estimate is
calculated on the basis of homogeneous risk groups instead of on a
policy-by-policy basis, the determination of the surrender strain as
defined in paragraph 3.134 may be burdensome and not necessary to
arrive at a sufficiently accurate capital
requirement.
Therefore, if it is proportionate to the nature,
scale and complexity of the risk, the comparison of surrender value
and best estimate provision might be made on the level of
homogeneous risk groups instead of a policy-by-policy
basis.
3.154. As the calculation on a homogeneous risk group
level is likely to result in the same or a lower capital requirement
than the policy-by-policy calculation, it seems necessary to set up
criteria for its application.
A calculation on the level of
homogeneous risk groups should be considered to be proportionate
if:
(a) the homogeneous risk groups appropriately distinguish
between policies of different lapse risk;
(b) the result of a
policy-by-policy calculation would not differ materially from a
calculation on homogeneous risk groups; and
(c) a
policy-by-policy calculation would be an undue burden compared to a
calculation on homogeneous risk groups which meet criteria (a) and
(b).
Factor-based formula for scenario effect
3.155.
For the two scenario calculations lapseshockdown and lapseshockup,
factor based simplifications were provided in QIS4.
These
simplifications attempt to approximate the effect of the permanent
change of lapse rates by
projection of the effect of a temporary
shock into the future.
The effect of a temporary change in
lapse rates can easily be measured by means of the surrender strain:
for example, the loss incurred in a portfolio with positive
surrender strain due to temporary change of lapse rates by 5
percentage points is approximately 5% of the surrender strain.
In order to for the permanence of the change in the
scenarios lapseshockdown and lapseshockup, this loss can be
multiplied with the duration of the portfolio in
question.
3.156. This approach results in formulas as
follows:
3.157. The
simplified calculation should be done at an appropriate
granularity.
3.158. The factor-based approximations should
only be used if they are proportionate to the nature, scale and
complexity of the risk.
In particular, as an application of
the scale criterion of the proportionality principle, the
simplification should only be used if the capital requirement for
lapse risk (determined with the simplification) is small compared to
the overall capital requirement.
A threshold of 5% of the
overall SCR (before adjustment for the loss-absorbing capacity of
technical provisions and deferred taxes) was tested in QIS4 and
appears to be appropriate.
Moreover, the simplification
should only be used, if the more sophisticated result of the
scenario analysis is not easily obtainable.
3.159. Although
the simplification was tested in QIS4, the results of the exercise
and the feedback could not clarify whether, firstly, the
simplification ensures a sufficiently high approximation quality
and, secondly, the approximation will be needed when Solvency II is
implemented.
Therefore, CEIOPS would appreciate feedback on
both of the following options:
Option 1: The standard formula
includes a factor-based simplification for lapse risk as defined
above.
Option 2: The standard formula does not include a
factor-based simplification for lapse risk.
In particular:
can we keep the factor based simplifications used in
QIS4?
3.2 CEIOPS’ advice
Treatment of lapse risk in
the scenario calculations
3.160. In the scenario calculations
of the SCR standard formula, the revaluation of technical provisions
should allow for relevant adverse changes in option take-up
behaviour of policyholders under the specified
scenario.
Scope of the lapse risk sub-module
3.161. In
relation to the policyholder options that the lapse sub-module
covers, a comprehensive approach should be taken. Ideally, the
module should take account of all legal or contractual policyholder
options which can significantly change the value of the future
cash-flows.
This includes options to fully or partly
terminate, decrease, restrict or suspend the insurance cover as well
as options which allow the full or partial establishment, renewal,
increase, extension or resumption of insurance cover.
3.162.
In the following, the term “lapse” is used to denote all these
policyholder options.
Calculation of the capital requirement
in the lapse risk sub-module
Simplifications
Calculation
on policy-by-policy basis
3.169. If it is proportionate to the
nature, scale and complexity of the risk, the comparison of
surrender value and best estimate provision referred to
in paragraph 3.164 might be made on the level of homogeneous risk
groups instead of a policy-by-policy basis.
A calculation on the
level of homogeneous risk groups should be considered to be
proportionate if
(a) the homogeneous risk groups appropriately
distinguish between policies of different lapse risk;
(b) the
result of a policy-by-policy calculation would not differ
materially from a calculation on homogeneous risk groups;
and
(c) a policy-by-policy calculation would be an undue burden
compared to a calculation on homogeneous risk groups which meet
criteria (a) and (b).
Factor-based formula for scenario
effect
3.170. Option 1: The standard formula includes a
factor-based simplification for lapse risk as defined
below.
Option 2: The standard formula does not include a
factor-based simplification for lapse risk.
3.171. A
simplified calculation of down Lapse and up Lapse as defined in
paragraph
3.172 may be made if the following conditions are
met:
(a) The simplified calculation is proportionate to nature,
scale and complexity of the risk.
(b) The capital
requirement for lapse risk under the simplified calculation is
less than 5% of the overall SCR before adjustment for the
loss-absorbing capacity of technical provisions and
deferred taxes.
For this comparison the overall SCR can be
calculated by means of the simplified calculation for the lapse
risk capital
requirement.
(c) The quantification of the
scenario effect defined in paragraphs 3.165 and 3.166 would be an
undue burden.
3.172. The simplified calculations are defined as
follows:
3.173. The
simplified calculation should be done at an appropriate
granularity.
Life Underwriting
Risk:
Introduction to Solvency ii Life Underwriting Risk
Solvency ii Mortality Risk
Solvency ii Longevity Risk
Solvency ii Disability Morbidity Risk
Solvency ii Life Expense Risk
Solvency ii Life Revision Risk
Solvency ii Lapse Risk
Solvency ii Life Catastrophe Risk
  
|
