Anomaly detection and data imputation within time series
IFRS17 Risk Adjustment Worked Example Part 2.pdf
1. I F R S 1 7 R I S K A D J U S T M E N T
F O R I N S U R A N C E C O N T R A C T S
P A R T 2
T A C K L I N G F U R T H E R A R E A S
SYED DANISH ALI
QUANTIFYING RESERVE RISK
WITH WORKED EXAMPLES
2. 1
2
3
4
Recap from Part 1
Diversification
IFRS17 Risk Adjustment for Reinsurance
IFRS17 Risk Adjustment for Liability for Remaining
Coverage
CONTENTS
I 2
3. 1
2
3
4
IFRS17 BEL and Stochastic Reserving
Final Notes
CONTENTS
IFRS17 Risk Adjustment Amortization
I 3
4. Executive Summary
In Part 1 of Risk Adjustment (RA) modeling for IFRS17, we undertook modeling on Cost of
Capital and Risk Metrics approach (VaR, TVaR, PHT) for ultimate view under IFRS17 as well
as from one-year view under Solvency 2. There were few further areas to develop in Part 1
that we have worked on in Part 2 which are 1) Diversification 2) RA for reinsurance 3) RA
for Liability for Remaining Coverage LRC and 4) amortization of RA. With this work done,
we have A to Z understanding of modeling for RA for general insurance products under
both GMM and PAA.
RM for Long-term life insurance is different under both GMM and VFA models and would
need to developed separately from this general insurance RA worked out here in Part 1
and 2. But this methodology can apply for group life and group credit too.
The main result is that RA for LIC (Liability for Incurred Claims) is leading to 14% increase
in reserves. Of course, this increase will be different for all other datasets such as at next
year valuation date but an increase will be there. The management must know that
IFRS17 RA will lead to increase in unpaid claim reserves and that it is not likely possible to
have same level of reserves under IFRS17 as are calculated under IFRS4. This will have
implication of decrease in net profits but if the increase in reserves are not followed by
the management by having zero RA, it will mean lack of compliance with IFRS17.
TVaR at 40% has been selected from multiple methodologies to represent RA for LIC. The
total impact for increase in reserves would be RA for LIC increase in unpaid claim reserves
and RA for LRC on onerous portions of UPR+PDR premium reserves. Mirrored Clayton
Copula has been selected as diversification method because of upper tail dependency.
The diversification reduced RA for LRC by 10%.
Gross IBNR as at Valuation Date IBNR Basis IFRS4 IFRS17 % change
Motor PaidClaims 969,519 1,110,250 15%
Medical IncurredClaims 207,751 228,674 10%
Property PaidClaims 12,665 15,149 20%
Liabilities IncurredClaims 77,202 93,556 21%
TotalBusiness 1,267,137 1,447,628 14%
Gross IBNR as at Valuation Date
Selected RA
Diversified for LIC
Methodology
Selected
Gross IBNR as per
IFRS4
Gross IBNR as per
IFRS17
Motor 140,731 TVaR at 40% 969,519 1,110,250
Medical 20,923 TVaR at 40% 207,751 228,674
Property 2,484 TVaR at 40% 12,665 15,149
Liabilities 16,354 TVaR at 40% 77,202 93,556
Total 180,492 - 1,267,137 1,447,628
Diversification Selected RA
Motor 156,368
Medical 23,248
Property 2,760
Liabilities 18,171
Simple Total 200,546
Undiversified Total 200,546
Diversified Total 184,833
Diversification amount 15,713
Diversification as % of simple total 8%
Undiversified Total 200,546
Diversified Total 180,492
Diversification amount 20,055
Diversification as % of simple total 10%
Selected Diversification factor 20,055
Mirrored
Clayton
Copula
Approach
Correlation
Matrix
Approach
Selected RA
5. Executive Summary
This means that ultimate view is mostly
higher than 1 year view because most
of the reserves develop within 1 year
(only CoC discounted is less than 100%
due to lowering liability by assuming
3% discount rate; otherwise, this
would've been much higher; risk metric
that focus more on the tail, TVaR and
PHT have slightly higher RA than VaR.
Motor Medical Property Liabilities Total (Simple Sum)
IFRS17 Mack Method 75% 154,975 24,762 1,879 12,986 194,601
IFRS17 VaR 75% RA 155,671 24,005 2,319 15,578 197,574
IFRS17 TVaR 40% RA 156,368 23,248 2,760 18,171 200,546
IFRS17 PHT 1.85 RA 168,877 26,270 3,174 18,898 217,219
IFRS17 CoC (Average discounted
reserves) 157,423 24,488 2,958 17,616 202,485
IFRS17 CoC (Average undiscounted
reserves) 175,456 27,293 3,297 19,634 225,681
IFRS17 CoC VaR (Discounted
Reserves) 99.5 Percentile 182,795 28,435 3,435 20,455 235,120
1 year Risk Margin Solvency2 CoC 162,145 25,223 3,047 18,144 208,560
1 year RM VaR 147,888 22,324 2,157 14,955 187,324
1 year RM TVaR 145,422 22,085 2,704 17,626 187,838
1 year RM PHT 155,367 23,906 2,856 17,575 199,704
VarR/TVaR/PHT corresponding
parameters to 1 year RM Solvency
2 CoC
VaR 73.5%, TVaR
36.3%, PHT 1.74
VaR 74.7%, TVaR
37.5%, PHT 1.76
VaR 75.5%, TVaR
43.3%, PHT 1.93
VaR 74.6%, TVaR
40.3%, PHT 1.81 N/A
VaR ultimate view / 1 year view 105% 108% 108% 104% 105%
TVaR ultimate view / 1 year view 108% 105% 102% 103% 107%
PHT ultimate view / 1 year view 109% 110% 111% 108% 109%
CoC (Average discounted reserves)
ultimate view/ 1 year view 97% 97% 97% 97% 97%
CoC (Average undiscounted
reserves) ultimate view/ 1 year
view 108% 108% 108% 108% 108%
CoC (VaR Discounted reserves
99.5th Percentile) ultimate view/ 1
year view 113% 113% 113% 113% 113%
RA for LIC Undiversified detailed results
Solvency 2-1
year view
IFRS17 -
Ultimate View -
Risk Metrics
IFRS17-
Ultimate View -
Cost of Capital
Ratios of
ultimate
views/1 year
view
6. Executive Summary
We undertook stochastic simulation as at valuation date for Best Estimate Liability instead
of only for Risk Adjustment as this is needed under IFRS17 instead of only deterministic
point-based estimates as currently determined under IFRS4.
RA can be amortized same as earning pattern of premium for LRC and on basis of run-off
of liabilities for LIC.
RA for reinsurance will be straightforward in case only there is only proportional
reinsurance arrangement and no other reinsurance arrangements. If it takes other than
quota arrangements in the future, RA for reinsurance will need to be worked out
separately from Gross and Reinsurance combined triangle can be made from 3 sub-
triangles 1) proportional reinsurance 2) non-proportional and 3) facultative. Same
methodologies can be applied to reinsurance although it will be more erratic and so
require higher percentile.
Claims can also be segregated into attrition and large claims different triangles but
methodology will remain the same.
Even if PAA is applicable to most of a company’s contracts, onerous portion of contracts
will require RA for LRC and so expecting 0 RA for LRC because PAA is applicable is a wrong
expectation to have. We have modeled RA for LRC on whole of UPR + PDR and that is
showing 26% increase in unearned reserves in IFRS17 compared to IFRS4. We don’t
require RA on 100% of LRC UPR+PDR and will require RA on only the onerous elements
and so this increase will be quite lower than 26%. RA for LRC will be 0 only if there are 0
onerous contracts, which is very unlikely to happen given how common cross-
subsidization commercial practices are.
IFRS4 UPR+PDR 18,228
IFRS17 LRC Under PAA (BEL only) 18,228
IFRS17 LRC Under GMM (BEL + RA) 23,032
RA for LRC 4,804
RA as % of BEL 26%
RA for LRC
7. 1 ) R E C A P F R O M P A R T 1
Revisiting previous work done and Overview of Part 2
8. 1) Recap of Part 1 and Overview of Part 2
In Part 1 of Risk Adjustment (RA) modeling for IFRS17, we undertook modeling on Cost of Capital and Risk Metrics
approach (VaR, TVaR, PHT) for ultimate view under IFRS17 as well as from one-year view under Solvency 2.
There were few further areas to develop in Part 1 that we have worked on in Part 2 which are 1) Diversification 2) RA for
reinsurance 3) RA for Liability for Remaining Coverage LRC and 4) amortization of RA. With this work done, we can have
A to Z understanding of modeling for RA for general insurance products under both GMM and PAA.
RM for Long-term life insurance is different under both GMM and VFA models and would need to developed separately
from this general insurance RA worked out here in Part 1 and 2.
Short term group life can be worked out just like general insurance under GMM and so can group credit life if opted for
IFRS17 and not IFRS9. There are considerable unique aspects to IFRS17 specific for Life insurance and they need to be
developed separately. Unit linked business goes to VFA and other long term traditional products like endowment, term
life, with-products for GMM. The investment portion of unit linked needs to go to IFRS9 and economic credit risk
modeling done on that portion. There is also scope for implementing Economic Scenario Generators (ESGs) like this one
done by SOA and AAA.
9. 1) Recap of Part 1 and Overview of Part 2
1 year view Vs ultimate view
Where: • R0 is the opening reserve, and is known
(i.e. VAR(R0) = 0); • P(t) are the payments during
calendar year t, solely for claims already happened
at the moment of evaluation t = 0 (i.e., it is assumed
that there is no new business); • R1 is the closing
reserve after having observed P(1). CDRt is the
Claim Development Result after t years from the
instant of evaluation, i.e. the difference between
actual and expected over the specified t th periodIn
summary, the “ultimate view” assesses all the
possible reserve paths1 until reserve run-off, whilst
the “one-year view” assesses only the different
paths over the first year and the resulting reserve an
actuary would estimate after observing each of
these one-year paths
Source: A Practitioner’s Introduction to Stochastic Reserving Alessandro Carrato MSc
FIA IOA, Gráinne McGuire PhD FIAA, Robert Scarth PhD 2016-04-21
12. 2) Diversification
Diversification Selected RA
Motor 156,368
Medical 23,248
Property 2,760
Liabilities 18,171
Simple Total 200,546
Undiversified Total 200,546
Diversified Total 184,833
Diversification amount 15,713
Diversification as % of simple total 8%
Correlation
Matrix
Approach
Selected RA
15. 2) Diversification
The mirrored Clayton Copula has upper tail dependence. This is a very important feature to have which Gaussian Copula
lacks. This means that variables are correlated but in extremes, they behave independently. This makes sense in a normal
market but not during market crisis and hence why copulas behaved poorly during 2008-9 Financial crisis as it massively
under-stated the probability of defaults when correlations became 1 for MBOs and CDOs. Another interesting and unique
example is joint life insurance long term product pricing where both spouses are insured together. Death of spouse
materially increases chances of death for the surviving spouse as well due to the ‘broken heart’ syndrome. If pricing doesn’t
recognize this correlation, it would be significantly underpriced.
For further intuitive understanding of Copulas, please refer to these videos by Paul Sweeting.
• Copulas: Learning the Basics Part 1https://www.youtube.com/watch?v=b6aZJuwE3Cs
• Copulas 2: After the Basics: https://www.youtube.com/watch?v=gzUxg0OUHU4
The mean and standard deviation for our 4 hypothetical lines of business to feed into the clayton copula model is:
The simulation is also shown below. The results on the right show that 5% of RA is decreasing due to allowing for
Diversification under Mirrored Clayton Copula Approach (Source: IAA Risk Adjustment Monograph). We should select this
approach over the simple Correlation Matrix as it is more technically sound.
Theta 1.50
Line X Mean 303,741
Line X Std Dev 60,181
Line Y Mean 31,122
Line Y Std Dev 10,667
Line Z Mean 1,588
Line Z Std Dev 1,037
Line AA Mean 11,806
Line AA Std Dev 6,676
Assumptions
Undiversified Total 200,546
Diversified Total 180,492
Diversification amount 20,055
Diversification as % of simple total 10%
Mirrored
Clayton
Copula
Approach
16. 3 ) I F R S 1 7 R I S K
A D J U S T M E N T F O R
R E I N S U R A N C E
Current Practice under IFRS4, detailed description of RA for Reinsurance in IFRS17
18. 4 ) I F R S 1 7 R I S K
A D J U S T M E N T F O R
L I A B I L I T Y F O R R E M A I N I N G
C O V E R A G E
Current Practice under IFRS4, detailed description of RA for LRC in IFRS17
21. 4) IFRS17 Risk Adjustment for LRC
Line of business STD Premium Risk STD Reserve Risk
Motor vehicle liability 10% 9.5%
Other motor 7% 10%
MAT 17% 14%
Fire 10% 11%
3rd party liability 15% 11%
Credit 21.5% 19%
Legal exp. 6.5% 9%
Assistance 5% 11%
Miscellaneous. 13% 15%
Medical expense 4% 10%
Income protection 8.5% 14%
Workers' compensation 5.5% 11%
Group Life insurance 17% 20%
Standard Deviation of Premium and Reserve Risk
1:Motor vehicle liability
2:Other motor 3:MAT 4:Fire 5:3rd party liability 6:Credit 7:Legalexp. 8:Assistance 9:Miscellaneous.
1:Motor vehicle liability 1 0.5 0.5 0.25 0.5 0.25 0.5 0.25 0.5
2:Other motor 0.5 1 0.25 0.25 0.25 0.25 0.5 0.5 0.5
3:MAT 0.5 0.25 1 0.25 0.25 0.25 0.25 0.5 0.5
4:Fire 0.25 0.25 0.25 1 0.25 0.25 0.25 0.5 0.5
5:3rd party liability 0.5 0.25 0.25 0.25 1 0.5 0.5 0.25 0.5
6:Credit 0.25 0.25 0.25 0.25 0.5 1 0.5 0.25 0.5
7:Legalexp. 0.5 0.5 0.25 0.25 0.5 0.5 1 0.25 0.5
8:Assistance 0.25 0.5 0.5 0.5 0.25 0.25 0.25 1 0.5
9:Miscellaneous. 0.5 0.5 0.5 0.5 0.5 0.5 0.5 0.5 1
Medical
expense
Income protection
Workers'
compensation
Medical expense 1 0.5 0.5
Income protection 0.5 1 0.5
Workers' compensation 0.5 0.5 1
RA for LRC 4,804
Function of the STD 5.8%
STD 8.9%
Total Volume measure 83,328
Confidence Interval 75.0%
Z-Factor 0.674
Diversified Total 7,429
IFRS4 UPR+PDR 18,228
IFRS17 LRC Under PAA (BEL only) 18,228
IFRS17 LRC Under GMM (BEL + RA) 23,032
RA for LRC 4,804
RA as % of BEL 26%
Undiversified Total 10,931
Diversified Total 7,429
Diversification Amount 3,503
Diversification % 32%
26% increase in IFRS17 GMM due to RA LRC than under
IFRS17 PAA.
Alternatively, we could ignore the reserve risk part and
only look at the premium risk part for RA for LRC as it can
be argued that BEL takes UPR and PDR into account
independently and only premium variability factor is
needed. Confidence interval can be increased from 75% to
95% then.
22. 5 ) I F R S 1 7 R I S K
A D J U S T M E N T
A M O R T I Z A T O N
Current Practice under IFRS4, detailed description of RA for LRC in IFRS17
24. 6 ) I F R S 1 7 B E S T E S T I M A T E
L I A B I L I T Y A n d n e e d f o r
S t o c h a s t i c R e s e r v i n g
Current Deterministic Reserving Practice under IFRS4, stochastic loss reserving need
for BEL in IFRS17 and not just RA
25. 6) IFRS17 Best Estimate Liability
https://www.casact.org/sites/default/files/2021-02/working-paper-ali2-2017-
08.pdf
29. 7 ) F I N A L N O T E S
1) Ending Note and Fur ther areas to develop in RA modeling
2) Key takeaways
3) More Key points
4) Lessons to live by
5) Recap: what we covered in this presentation
30. Ending Note and Further areas to develop
In this presentation, various concepts associated with the quantification of reserve risk have been connected. The analytic formula-based approaches of
Mack for the lifetime view of reserve risk, and Merz and Wuthrich for the one-year view of Solvency II, have been compared to simulation-based results
obtained by bootstrapping Mack’s model, supplemented with the re-reserving approach. Furthermore, the lifetime and one-year views were brought
together by considering a sequence of one-year views until the liabilities are extinguished. Again, this was considered analytically, using Merz and
Wuthrich, and using a simulation-based approach by applying re-reserving recursively.
IFRS 17 risk adjustments are also required on a gross and reinsurance basis. Clearly, it is the net position that is most relevant for the interpretation of an
insurance entity’s financial position, so it seems appropriate to estimate risk adjustments from distributions of gross and net discounted fulfilment cash-
flows, then taking the difference as the reinsurance risk adjustment. Reinsurance modelling to obtain an accurate distribution of the net discounted
fulfilment cashflows (together with an assessment of credit risk) could be complex. In particular, the current actuarial practice of applying an approximate
net-to-gross ratio looks increasingly inadequate (where non-proportional reinsurance treaties exist), and triangle methods for attritional claims may need
to be supplemented by individual claims modelling for large claims, with accurate reinsurance modelling. Furthermore, risk adjustments are required for
groups of contracts, not just at the aggregate entity level (or holding company level for a multinational group), which raises questions about allocation of
risk and diversification. a simulation framework can be used (using copulae to apply dependencies when aggregating), but the issues are complex.
If the cost-of-capital technique is used for IFRS 17 risk adjustments, it should be recognized that this will be different from a Solvency II risk margin.
Solvency II considers the one-year view of risk for capital requirements, whereas the lifetime view of risk is more appropriate under IFRS 17. A distribution
of the remaining total cash-flows at each future time period is more appropriate as a basis for estimating capital requirements (although as discussed in
section 6 and Appendix 3, the time perspective becomes important). Furthermore, cost-of-capital and discount rates are entity specific under IFRS 17 but
prescribed under Solvency II. The cost of-capital technique is considerably more complex than simply applying a risk measure to a distribution of fulfilment
cash-flows, and requires more parameters to select and justify; it requires an opening capital requirement, future capital requirements, a cost-of-capital
rate and a yield curve for discounting. Since the equivalent “confidence level” is required anyway under IFRS 17, it questions why the cost-of-capital
method would be used at all. A distribution of discounted fulfilment cash-flows is required for the equivalent confidence level, so it seems more
straightforward to calculate IFRS 17 risk adjustments simply from a risk measure applied to that distribution. Given the distribution, the only input to select
is the entity specific risk tolerance level.
30
31. More Key Points
All risks need to be measurable and be
quantified.
Measurable
No use if calculating Risk Adjustment needs an
unreachable budget for the company. But
reasonable budgets should also be there otherwise
patchworks can mean quality can suffer. The person
who buys expensive cries once but the person who
buys cheap cries ten times.
Cost effective
The RA needs to work across very diverse lines of
business including motor, medical, short-term life,
long-term life, marine, engineering, liabilities and
so on.
Multiple lines
32. Lessons to Live by
Pragmatic Vision and
Budgets
Leadership
Quality
Deep Expertise
IFRS17 is unlike normal work like
reserving or pricing which actuaries
have repeated thousand of times.
This is being done for the first time
worldwide and no one has done A to
Z all of it before so it’s better to
over- prepare than under-prepare as
the consequences of under-
preparation are far worse than of
over-preparation. That vision needs
to be backed up by reasonable
budgets. Going for unreasonably low
budgets mean lots of pain
afterwards.
Deep expertise is needed in
order to implement solutions
that are technically sound in line
with principles of IFRS17 instead
of simple patchworks.
The binary view that
insurer is compliant with
IFRS17 or not compliant is
misleading as quality of
compliance differs
drastically across different
insurers and markets.
Unless the top management of
company takes IFRS17 seriously,
implementation will suffer
drastically. It has been noticed
across various markets that 90%
or more work is done by
consultants but there is extremely
low ownership and knowledge by
company employees of IFRS17.
Collaboration
Across different
segments of business
from Finance to
underwriting to IT and
Actuaries is crucial
Communication
tailored to specific
stakeholders is key
Collaboration
Communication
33. 01 02 03
04 05 06
Recap - What we covered in this presentation
Modeling for Correlation Matrix and
Copulas.
Diversification
Current practice under IFRS4.
Detailed description of RA
requirements for Reinsurance under
IFRS17.
IFRS17 Risk Adjustment for
Reinsurance
Quantifying premium variability.
IFRS17 Risk Adjustment
for LRC
Amortization methodologies.
IFRS17 RA Amortization
IFRS17 Best Estimate Liability and the
need for stochastic reserving and
moving on from point based
deterministic reserving as done
currently.
IFRS17 BEL
Key notes to keep in view.
Final Notes
34. T H A N K Y O U !
A n y Q u e s t i o n s ?
SYED DANISH ALI