1 philipp keller, federal office of private insurance brussels, 14 october 2005 swiss solvency test
Post on 19-Dec-2015
218 views
TRANSCRIPT
1
Philipp Keller, Federal Office of Private Insurance Brussels, 14 October 2005
Swiss Solvency Test
2
Contents
•Designing a Solvency System– Basic Requirements– Principles vs Rules
•Concept of the Swiss Solvency Test•The SST Standard Model•Experiences from the Field Tests• Internal Models•Group Aspects
3
Designing a Solvency Test
Basic requirements of a risk-based solvency system:
•Requirements and definitions have to follow from regulatory purpose (e.g. what is the risk margin for, what purpose does the solvency capital requirement (SCR) have, etc.)
• If internal models are to be used, the solvency system should be defined by underlying principles
•Building blocks have to fit together (SCR has to be linked to risk margin to avoid double counting of risks, the valuation of assets and liabilities has to be consistent, …)
• If the solvency system has to be embedded within insurance companies, it has to be founded on economic principles
•Avoid nontransparent mix of quantification of risks, limited eligibility of capital, limits and implicit safety margins which would make interpretation of SCR impossible
4
Designing a Solvency Test
A risk based solvency system has to address the needs of different stakeholders:
• Regulators: policyholder protection, setting capital requirements, incentive for risk management, obtaining a view on companies’ risk culture and risk awareness, achieving a competitive market place, …
• Policyholders, Investors, Shareholders, Brokers,…: Comparison of the financial situation of different companies, information on the efficiency of risk management, ALM etc.
• Insurers (CEO, CRO, CFO, Actuaries): Comparison with peers, use of internal models, analysis of contribution of different risk types, …
A risk based solvency system has many purposes and needs to be a reasonable compromise to satisfy the needs of many stakeholders
5
Designing a Solvency Test
Definition of Capital
Valuation
Time Horizon
Risks to Quantify
Risk Measure
Definition of Ruin
Define risk typology, decide which risks are quantified and which treated qualitatively
Define risk margin in SST: risk margin = lowest acceptable level of risk bearing capital
Decide over which time horizon risk capital needs to cover risks
Define valuation methodology of assets and liabilities
Define risk bearing capital, e.g. value of assets less best-estimate of liabilities, define which forms of capital are eligible
Define measurement of risks, e.g. Value at Risk, Expected Shortfall, etc.
Operational Implementation
In the last stage, define standard model (if necessary), details of the system, e.g. correlations vs copulas, etc.
6
Implications of Principles vs Rules
Principle-based
Rule-based
Principle-based standards describe the objective sought in general terms and require interpretation according to the circumstance.
Objective
Objective
Risk Based Capital Requirement
Companies tailor approach such that clearly stated objective is attained
Rule based approach does not allow truly company specific risk assessment
Attained result depends on how well rules capture the situation of the insurer
Objective can be attained if companies interpret principles faithfully. The objective is defined by a (theoretically) correct quantification and allocation of group diversification
=
7
Contents
•Designing a Solvency System•Concept of the Swiss Solvency Test
– Principles– Quantified Risks– Risk Measure– Valuation– Risk Margin– Change in Risk Bearing Capital– Capital
•The SST Standard Model•Experiences from the Field Tests• Internal Models•Group Aspects
8
Principles Definitions
Guidelines
• Principles define concisely the objectives
• Definition of terms and concepts so that meaning and possible interpretation of principles become clear
Glossary
The SST is defined not by the Standard Model but by underlying principles
Core of the Solvency Test
Standard Model
• Guidelines help in interpretation
• Standard Model allows use of Solvency Test also by small companies
The SST Concept: Principle-Based
The more laws and order are made prominent, the more thieves and robbers there will be, Lao-tzu
9
Principles defining the SST
1.All assets and liabilities are valued market consistently
2.Risks considered are market, credit and insurance risks
3.Risk-bearing capital is defined as the difference of the market consistent value of assets less the market consistent value of liabilities plus the risk margin
4.Target capital is defined as the sum of the Expected Shortfall of change of risk-bearing capital within one year at the 99% confidence level plus the risk margin
5.The SST defines an insurer’s capital adequacy if its target capital is less than its risk bearing capital
6.The scope of SST is legal entity and group / conglomerate level domiciled in Switzerland
7.Scenarios defined by the regulator as well as company specific scenarios have to be evaluated and, if relevant, aggregated within the target capital calculation
8.All relevant probabilistic states have to be modeled probabilistically
9.Partial and full internal models can and should be used
10.The internal model has to be integrated into the core processes within the company
11.SST Report to supervisor such that a knowledgeable 3rd party can understand the results
12.Disclosure of methodology of internal model such that a knowledgeable 3rd party can get a reasonably good impression on methodology and design decisions
13.Senior Management is responsible for adherence to principles
Defines Output
Defines How-to
Transpar-ency
Responsi-bility
10
Risk Measures: Expected Shortfall
Expected Shortfall is a coherent risk measure
Shareholder: Only default or non-default is relevant not how bad the state of the insurer is in case of default as shareholders have a put-option on the insurer
(Merton) Value-at-Risk is appropriate
Policy Holder: In case of default, it matters how much capital is
left Expected Shortfall is more appropriate than VAR
From the perspective of an insurance regulator, Expected Shortfall has advantages compared to Value at Risk
For an insurer, Expected Shortfall has advantage of being coherent:
•Allocation of risk and risk management of subunits is possible
•ES is easier to explain to management:
- ES1%=average one-in-a-hundred-years loss
- VaR1% = the loss that is in 99-out-of-a-100-years not exceeded
The Expected Shortfall of a random variable X to the confidence level 1- (ES) is given by
ES[X] =1/ · E[ max( X- VaR[X], 0 )] + VaR[X]
11
The SST Concept: The economic view
How to measure risks?
• Accounting risk or economic risk?
Reported earnings follow the rules and principles of accounting. The results do not always create measures consistent with underlying economics. However, corporate management’s performance is generally measured by accounting income, not underlying economics. Therefore, risk management strategies are directed at accounting, rather than economic performance.
Enron in-house risk-management handbook
For a risk-based solvency system, risks need to be measured objectively and consistently → economic risk rather than accounting risk
→ Market Consistent Valuation of Assets and Liabilities
12
The SST Concept: The economic view
Market Consistent
Assets Liabilities
Best-Estimate Provisions
Risk Margin
Market consistent provisions
Risk bearing capital
Wherever possible, market-consistent valuation is based on observable market prices (marking to market)
If such values are not available, a market-consistent value is determined by examining comparable market values, taking account of liquidity and other product-specific features, or on a model basis (marking to model)
Market-consistent means that up to date values are used for all parameters
Best-estimate = Expected value of liabilities, taking into account all up to date information from financial market and from insurance.
All relevant options and guarantees have to be valued.
No explicit or implicit margins
Discounting with risk-free interest rate
Risk Margin for inherent risk in liability portfolio
Valuation of policyholder-options: Assume realistic behavior of policy holders, but option exercise depends on financial market parameters
One approach to value options is using replicating portfolio of traded financial instruments
13
The SST Concept: The SST in a Nutshell
TC = ES[ΔRBC] + RM
Risk MarginChange of Risk-Bearing Capital
Target Capital
Expected Shortfall
Covers risks emanating during a one-year time horizon
Risk margin should be sufficient to allow a run-off or portfolio transfer
14
The SST Concept: Risk Margin
Risk Margin: to cover policyholders against risks emanating beyond 1 year
Risk Margin
Best Estimate
Provisions
Possible Approaches:
• Statutory: Taking undiscounted reserves, using prudent assumptions, adding a simple factor on best-estimate etc.
• Quantile: Taking e.g. the 75% quantile of the ultimate loss distribution of the liabilities: Used by APRA for P&C liabilities, discussed within Solvency 2.
• SCR / ES: To cover risks which emanate during a 1-year time horizon
• Risk Margin: To cover risks during the whole run-off of the portfolio
• Market Value Margin: the additional amount on top of the best estimate which is required by a willing buyer in an arms-length transaction to assume the liabilities the loss reserves are held to meet: Discussed within Fair Value Accounting, used (partly) within the SST.
The Risk Margin and SCR/ES cover different risks:
There should not be double-counting between SCR and Risk Margin
15
The SST Concept: Risk Margin
Definition: The risk margin is the smallest amount of capital which is necessary in addition to the best-estimate of the liabilities, so that a buyer would be willing to take over the portfolio of assets and liabilities.
Idea: A buyer (or a run-off company) needs to put up regulatory capital during the run-off period of the portfolio of assets and liabilities
a potential buyer needs to be compensated for the cost of having to put up regulatory capital
Risk Margin = cost of capital of the present value of future regulatory risk capital associated with the portfolio of assets and liabilities
Problem: How to determine future regulatory capital requirement during the run-off of the portfolio of assets and liabilities?
-> Assumptions on the evolution of the asset portfolio are necessary
16
The SST Concept: Risk Margin
Key Idea:
Liabilities: Assume no new business
Assets: Assume that initial asset portfolio is rebalanced such that it matches optimally the liabilities. The speed of the rebalancing is constrained by liquidity of assets (it takes longer to liquidate for real estate than for government bonds). The time until the optimal replicating asset portfolio is achieved depends on the asset mix.
• The insurer setting up the risk margin should not be penalized if, after the transfer, the insurer taking over the portfolio does not minimize the regulatory risk capital requirements as fast as possible.
• The insurer taking over the portfolio of assets and liabilities should be compensated if the insurer setting up the risk margin invested in an illiquid asset portfolio.
17
ES with optimally replicating asset portfolio
ES with portfolio converging from actual to replicating portfolio taking into account illiquidity of assets Sequence of Achievable Replicating Portfolios
Years
ES: 1-Period (e.g. 1 year) risk capital = Expected Shortfall of risk-bearing capital
t=1 t=2 t=3
Achievable Replicating Portfolio has converged to Replicating Portfolio
The SST Concept: Risk Margin
t=0
ES at t=0 does not enter calculation of the risk margin necessary at t=0 risks taken into account for 1-year risk capital and risk margin are completely disjoint and there is no double-counting
Future ES entering calculation of risk margin at t=0
t 2
TC=ES ΔRBC(1) +sp ES ΔRBC(t)
18
Risk Margin
0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.080
0.1
0.2
0.3
0.4
0.5
0.6
0.7
Risk Margin / Best Estimate
Ris
k M
argi
n /
1 Y
ear
Ris
k C
apita
l
Risk Margin / Best Estimate vs Risk Margin / 1-Year Risk Capital
Nonlife
Life
Risk Margin / Best Estimate vs Risk Margin / ES[RBC], based on provisional data of Field Test 2005
Life companies writing predominately risk products
Life companies writing predominately savings products
19
Change in Risk Bearing Capital
Assets Liabilities
RBC(0) RBC(1)
RBC(0) should be such that at the end of the year, even when a large loss with P<1% occurs, the insurer‘s available RBC covers (on average) still the risk margin
Investment profit (above risk-free) + known payoffs (deterministic)
New business during one year (deterministic)
Changes in value of assets (stochastic)
Changes in value of liabilities + claims during 1 year (stochastic)
Assets Liabilities
Year 0 Year 1
State 1.1 known / deterministic
State 31.12 unknown / stochastic
The SST requires the quantification of the randomness of risk bearing capital in one year (the probability distribution of RBC). From this follows the determination of target capital
20
Change in Risk Bearing Capital
(0)1
(0)1
( (0) )1
IE R rA UPR P K
r
(1)
(0)1
( )1
CY
CY
E DP K E S
r
)0()0(
1
)1()1(
)0(1
)1()1(
)0(1 1
][][
1
][))()0((
1
][PY
PYPYCY
CYCYII Rr
DEDSE
r
DEDKUPRPA
r
RER
RBC(0)RBC(1)
Year 0 Year 1
)0()0()0(
1
)1(
)0(1
)1(
11PYPYPY
PYCYCY
CY RdCr
DESES
r
DE
21
Change in Risk Bearing Capital
Expected asset return over risk-free
Insurance risk (deviation of technical result from expectation)
Expected insurance (technical) result
(0)1
(1)(0)
1RBC
RBCr
CYCY SEr
DEKP
)0(1
)1(
1)(
(1) (1)
(0) (0)(0) (0)
1 11 1CY PY
CY CY PY PY PY
E D E DS E S C d R
r r
(0)1
(0)1
( (0) )1
IE R rA UPR P K
r
(1) (1) (1) (1)(0)
(0) (0) (0)1 1 1
[ ][ ] [ ]( (0) ( ) ) [ ]
1 1 1CY CYI I PY PY
CY PY
D E DR E R D E DA P UPR K E S R
r r r
ALM risk
Current Year Risk
Previous Year Risk
22
Expected Shortfall of Risk Bearing Capital
Definition. An insurer satisfies the SST when:
ES[RBC(1) 0]≥ rm;
where rm denotes the risk margin.
RBC(1) = (rbc(0) + r(0) + p - k)(1 + RI ) - S(1) - R(1)
RBC(1) in function of terms known at t=0:
rbc(0)=a(0) - R(0): Risk-bearing capital at t=0, a(0): assets at t=0, R(0): liabilities at t=0 (best-estimate)
p: Expected premium during [0,1] r(0): Liabilities at t=0
(Current year) r0: risk-free rate
k: Expected costs during current year upr: unearned premium reserve
S(1): Claim payments during current year RI: Asset return
R(1): Liabilities at t=1 Notation simplified
23
RBC(1) = (rbc(0) +r (0) + p - k)(1 + RI ) - S(1) - R(1)
ES[RBC(1) 0]=ES[(rbc(0) + r(0) + p - k)(1+RI) - S(1) - R(1)]
= ES[(rbc(0) + r(0) + p - k)(1 + r0-r0+RI) - S(1) - R(1)]
rbc(0)(1+r0)+ES[r(0)+p-k)(1+r0-r0+RI )
+rbc(0)(RI-r0)-S(1)-R(1)] ≥rm
rbc(0) ≥-ES[(r(0)+p-k)(1+r0-r0+RI )
+rbc(0)(I-r0)-S(1)-R(1)] ≥rm
0 0
(a(0)-upr+p-k)(1+I)-S(1)-R(1) rmrbc(0) -ES -(a(0)-upr+r(0)) +
1+r 1+r
Expected Shortfall of Risk Bearing Capital
24
Contents
•Designing a Solvency System•Concept of the Swiss Solvency Test•The SST Standard Model
– Market Risk– Nonlife Risk
• Current Year Risk• Previous Year Risk
– Life Risk– Scenarios
•Experiences from the Field Tests• Internal Models•Group Aspects
25
The SST Concept: General Framework
SST Concept
Models Scenarios
Aggregation Method
Fin
anci
al
Ris
ks
Cre
dit
R
isks
Insu
rance
R
isks
Standard Models or Internal Models
Mix of predefined and company specific scenarios
Asset-Liability Model
Target Capital SST Report
26
The SST Concept: Standard Models
•Market Risk–For many companies this is the most important risk (up to 80% of total target capital emanating from market risk)
–Needs to be modeled with particular care
–Most relevant are interest rate risk, real estate risk, spread risk, equity risk
–Market risk model needs to take into account ALM
•Credit Risk–Credit risk is becoming more important as companies go out of equity and into corporate bonds
–Many smaller and mid-sized companies do not yet have much experience in modeling credit risk
•Insurance Risk (Life)–For many life companies with predominantly savings product, pure life insurance risk is not too important
–Life insurance risk is substantial for companies selling more risk products / disability
–Model needs to capture optionalities and policyholder behavior
•Insurance Risk (Nonlife)–Premium-, reserving- and cat risk are important
–A broad consensus on modeling exists among actuaries
More information under:
www.sav-ausbildung.ch
27
Scenarios
•Historical – Share crash (1987)– Nikkei crash (1990)– European FX-crisis (1992)– US i.r. crisis (1994)– Russia crisis / LTCM (1998)– Share crash (2000/2001)
•Default of Reinsurer•Financial Distress
– Equity drop– Lapse = 25%– New business = -75%
•DeflationFor SST, RiskMetrics type model with given risk factors and associated volatilities and correlation matrix is used together with scenarios
Financial market risk often dominates for insurers adequate modeling of interest rate-, equity-,.. risks is key
Interest rate risk can not be captured solely by a duration number
Financial instruments have to be segmented sufficiently fine else arbitrage opportunities might be created
Regulatory requirements shouldn’t force companies to disinvest totally from certain investment classes (e.g. shares, private equity)
Standard Model: Market Risk
28
Standard Model: Market Risk
CHFEUR
USDGBP
Spreads•AAA•AA•A•BAA
Equity
•Shares•CHF•EUM•USD•GPB•JPY
•Real Estate•IAZI•Commercial•Rüd Blass•WUPIX A
•Hedge Funds
•Private Equityi.r. time buckets:1,…,10, 15,20,30+
75 Risk Factors:•4*13 interest rate•4 spreads•4 FX•5 shares•4 real estate•1 hedge fund•1 private equity•1 participations•3 implied volas
FX•EUR•USD •GBP•JPY
29
Asset Cash Flows
Liability Cash Flows
Year
Year
Netto Cash Flows A-L
Present Value von Asset - Liabilities
Change of present value of net cash flow (assets-liabilities) due to change in the 2 year CHF yield
The SST Concept: Cash Flow Based
0
1
2
3
CHF Yield Curve
Stressed 2Y Yield
Example: Sensitivity to 2 Year CHF Yield
- =
30
Standard Model: Nonlife
Normal claims Large claims
Compound Poisson
For each LoB, Pareto distribution with specified or company specific parameters
For each LoB, moments are derived by parameter- and stochastic risks (coefficients of variation)
Method of Moments with prescribed correlation matrix
Method of Moments with prescribed correlation matrix Lognormal
Reserving Risk
Discounted cash flows
Assets: bonds, equity, …
CF -> i.r. sensitivities
Asset Model: Covariance/Riskmetrics
approach
Normal
Premium risk Market risk
… …
…
Further aggregation with scenarios
Aggregation by Convolution
Lin
es
of
Bu
sin
ess
Aggregation by Convolution
First two moments of premium risk (normal claims) and reserving risk are aggregated using correlation -> two moments defining lognormal
31
Standard Model: Insurance Part
(0) (0) (0)( 1)CY CY PY PY PYP K d S d C R
Assumed deterministi
c
Current Year Risk Previous Year Risk
Claims which occur during 1 year:
Within each Line of Business:
• Normal claims and large claims
• Catastrophes which affect different LoBs simultaneously
Reserving risks due to:
• Randomness (stochastic risk)
• Reassessment of reserves (parameter risk)
32
Standard Model: CY Risk Normal Claims
Coefficient of Variation for a single claim in LoB i
22 param 2i i ,2
due to Parameter Riskdue to Stochastic Risk
Var[ ] 1CoeffVar CoeffVar ( CoeffVar( ) 1)
E[ ]
Ni
i jNi i
CY
C
Normal Claims: “High frequency claims”, different for each LoB.
Split Normal / Large claims is in standard-model defined, companies can adjust to their specific situation
For each LoB:
•Estimate Parameter & Stochastic Risk due to normal claims
•Then aggregate using correlation matrix ( first two moments define a Gamma distribution for normal claims )
For each LoB i:
Expected number of claims Yi,j in LoB j
, , ,1 , 1
2
i , , , j ,1 , 1
Var Cov ,
CoeffVar E (CoeffVar E ) (CoeffVar E )
NS NS NS NSCY CY i CY i CY j
i i ji j
NS NS NSCY i i j i CY i CY j
i i ji j
Var S S S S
S S S
33
Standard Model: CY Risk Normal Claims
Within the Standard Model: CoeffVari for parameter risk and coeffvar(Yi,j) for single normal claims for two different splits normal / large claims is given
Parameter RiskLoB i CoeffVari CoeffVar(Yi,j) CoeffVar(Yi,j)
Split at CHF 1 Mio Split at CHF 5 MioMFH 3.50% 7.00 10.00MFK 3.50% 2.50 2.50Sach 3.00% 5.00 8.00Haftpflicht 4.50% 8.00 11.00UVG 3.50% 7.50 9.50Unfall ohne UVG 3.00% 4.50 5.50Kollektiv Kranken 3.00% 2.50 2.50Einzelkranken 5.00% 2.25 2.25Transport 5.00% 6.50 7.00Luftfahrt 5.00% 2.50 3.00Finanz und Kaution 5.00% 5.00 5.00Other 5.00% 5.00 5.00
Stochastic RiskFor each LoB, each
company has to:
• Define a split into normal and large claims
• Estimate the expected number of normal claims
• Estimate the expected normal claim amount
• If split normal/large is 1 or 5 Mio, then standard values can be used
34
Standard Model: CY Risk Large Claims
Large claims: Large single claims and accumulation of claims due to a single event
For each LoB j: Total amount due to large claims is modeled as Compound Poisson with single claims Yi,j being Pareto distributed and number of claim Nj being Poisson. Then
, ,1
iN
LC LCCY i i j
j
S Y
Further assumption: SCYj are independent Total amount due to large claims over all LoB is again Compound Poisson
35
Standard Model: CY Risk Large Claims
.1
,0
)()( ,,
yy
y
yYPyF i
GSji
GSjiY
Pareto Distribution:
Smallest large claim comsidered: β threshold parameter
Shape parameter α: The smaller , the more heavy-tailed. If ≤k, k-th moment does not exist anymore
Table with standard parameters for companies lacking sufficient data
For numerical calculations the cut-off point of the distribution is very important
LoB b=1 Mio b=5 Mio Cut off PointMFH 2.50 2.80 I llimitéMFK Marktet Share * 1.5 Mrd. CHFSach 1.40 1.50 Company specific estimation of largest possible claimHaftpflicht 1.80 2.00 Company specific estimation of largest possible claimUVG inkl UVGZ 2.00 2.00 I llimitéUnfall ohne UVG CHF 50 MioKollektiv Kranken 3.00 3.00 Company specific estimation of largest possible claimEinzelkranken 3.00 3.00 Company specific estimation of largest possible claimTransport 1.50 1.50 2 * largest possible sum at riskLuftfahrt not modeled as mostly reinsured in poolFinanz und Kaution 0.75 0.75 Company specific estimation of largest possible claimAndere 1.50 1.50 Company specific estimation of largest possible claim
α
36
Current Year Risk: Large Claims
1
~ ( ), ~ , , then is compund Poisson, ~ ( , )N
i Y i Yi
N Pois Y F iid S Y S F
Aggregation: Use the property that for each LoB, the loss S is compound Poison distributed
Let ~ ( , ), indep. Then ~ , is also compound Poisson.j
Yj
j Yj
j j j jj j j
j
FS F Z S
For Compound Poisson distributions, efficient algorithms exist to numerically evaluate the distribution function, e.g. Panjer Recursion
Consider the class of claim frequency distributions N with pn/pn-1=a+b/n, n=1,2,3,.. Where pn=P (N=n). For this class with claim sizes Y defined on the positive integers, the following recursive formula for the distribution of total claims holds:
1
( ) ( ) ( ), 1,2,3,...y
S Y Sj
jf y a b f j f y j y
y
See Insurance Risk Models, Panjer&Willmot
37
Standard Model: Previous Year Risk
Assumption: CPY*RPY(0) lognormal, with
expectation RPY(0) E[CPY]=1
Randomness of CPY due to parameter and stochastic risk
Stochastic Risk: due to randomness of single claims company specific estimation from historical run-off result. Determine for each LoB (where all historical data is on best-estimate basis)
Parameter Risk: Estimates of parameters are uncertain which affect all provisions of a LoB level of total provisions incorrectly chosen.
PY Risk: Reserving Risk, due to uncertainty of run-off result
For parameter risk: predefined values are given
For stochastic risk: company has to determine based on experience, e.g. using Mack
Parameter Risk: Parameters given for standard model:
MFH 3.50%MFK 3.50%Sach 3.00%Haftpflicht 4.50%UVG 3.50%Unfall ohne UVG 3.00%Kollektiv Kranken 3.00%Einzelkranken 5.00%Transport 5.00%Luftfahrt 5.00%Finanz und Kaution 5.00%Andere 5.00%
38
Volatility: Describes changes of risk factors within one year due to parameter-uncertainty
Stochastic risk will be included using company specific data if relevant
The volatilities have been set during discussions with specialist and represent a best-guess
Standard Model: Life
Assumptions: The risk factors are normal distributed with given volatilities. Change of risk bearing capital in function of the risk factors is linear The distribution over all risk factors is again (multivariate) normal distributed
Risk Factors: Volatility
Indiv. Group
•Mortality 5% 5%
•Longevity (trend) 10% 10%
•Disability 10% 20%
•Reactivation 10% 10%
•Lapse 25% 25%
•Option Exercise 10% 10%
39
Standard Model: Life
Correlations between risk factors for field test 2005:
Split into individual and group business. Full correlation between individual and group business risk factors, except for lapse
Einzel Gruppen
Ste
rblic
hkei
t
Lan
gleb
igke
it
i(x)
r(x)
Sto
rno
Opt
ions
ausü
bung
Ste
rblic
hkei
t
Lan
gleb
igke
it
i(x)
r(x)
Sto
rno
Opt
ions
ausü
bung
Sterblichkeit 1 0 0 0 0 0 1Langlebigkeit 0 1 0 0 0 0 1i(x) 0 0 1 0 0 0 1r(x) 0 0 0 1 0 0 1Storno 0 0 0 0 1 0.75 1 0.5Optionsausübung 0 0 0 0 0.75 1 0.5 1Sterblichkeit 1 1 0 0 0 0 0Langlebigkeit 1 0 1 0 0 0 0i(x) 1 0 0 1 0 0 0r(x) 1 0 0 0 1 0 0Storno 1 0.5 0 0 0 0 1 0.75Optionsausübung 0.5 1 0 0 0 0 0.75 1
40
Standard Model: Life
The model is simple and transparent: the company has to determine sensitivities with respect to life insurance risk factors and then can use correlation matrix and volatilities to arrive at distribution for life insurance risk
The normality assumption allows easy aggregation with market risk
,
,,
,
M IG
I GTM IG T
I G
M C
I CC
C G
Correlations between market risk factors and insurance risk factors (individual and group business)
For field test 2005, CM,IG=0, but in future correlation between market risk and insurance risk can easily be included (e.g. correlation between lapse and interest rate)
Correlations between market risk factors
Correlations between life insurance risk factors (within individual (I), within group (G) and between individual and group CI,G
41
The SST Concept: Scenarios
Scenarios can be seen as thought experiments about possible future states of the world. Scenarios are not forecasts, in that they need not predict the future development, but rather should illuminate possible but perhaps extreme situations. Scenarios are also different from sensitivity analysis where the impact of a (small) change of a single variable is evaluated.
Current state of the world
Alternate states of the world
“Ersatz experience is a better guide to the future than the real past and present”, Hermann Kahn in On Thermonuclear War
42
Standard Model: Aggregation with Scenarios
00 0
( ) ( ) ( )n n
i i i ii i
F x p F x p F x c
Assumption: During normal years, analytical models without scenarios are valid (described by a probability distribution F0), during exceptional years, additional losses due to events described in scenarios occur, causing a shift in the ‘pre-scenario’ distribution.
Scenario i, i=1,…,n occurs with probability pi and causes an additional loss of ci (ci<0)
The probability of a normal year is p0=1-p1-…-pn
The probability distribution of the risk during a year with aggregated scenarios is:
F(x) is the weighted mean of shifted distributions F0(.-ci), i=1,…,n, where c0=0.
43
The SST Concept: Scenarios
Historical Scenarios: Stock Market Crash 1987, Nikkei Crash 1989, European Currency Crisis 1992, US Interest Rates 1994, Russia / LTCM 1998, Stock Market Crash 2000
Financial Distress: Increase of i.r., lapse, no new business, downgrading of company,…
Deflation: decrease of i.r.
Pandemic: Flu Pandemic with given parameters (e.g. number of death, sick-days, etc.)
Longevity
Reserving: Provisions have to be increased by 10%
Hail (Swiss specific): Given footprints
Default of Reinsurer: Reinsurer to which most business has been ceded defaults
Industrial Accident: Accident at chemical plant
Personal Accident: large accident during company outing or mass panic in soccer stadium
Anti-selection for Health Insurers: all insured with age < 45 lapse
Collapse of a dam (Swiss specific)
Terrorism
Global Scenarios (for groups&reinsurers)
Property Cats (earthquake, windstorm)
Special Line Cats: Aviation (2 planes collide, marine event, energy event, credit&surety event
44
Contents
•Designing a Solvency System•Concept of the Swiss Solvency Test•The SST Standard Model•Experiences from the Field Tests
– Experiences– Comments– Some Results
• Internal Models•Group Aspects
45
Results of Field Test
•It is a challenge to stay principle-based, since explicit rules are desired by some of those who have to implement the SST
•The possibility of analyzing the contributions of different risks to required capital are seen as a big advantage in particular for companies not yet using a full internal model
•A risk based solvency framework entails close cooperation and communication of different sections within insurance companies
•Substantial simplification are not perceived to be feasible if explanatory power of SST is to be kept
•Solvency 1 (statutory view) and SST are not yet compatible Solvency 1 will have to be made more consistent so as not to send out conflicting signals
•Modelling of participations and contingent risk and capital transfer solutions will be challenging
46
Impressions from the Industry
SST will favour large companies that have already sophisticated risk-based management systems in place …’
‘Small companies without internal model will be punished by the Standard Approach of SST…’
‘SST may call for a complete overhaul of risk management …’
‘Technical implementation can become a problem …’
‘… transparency and fair values will further increase the volatility of earnings …’
‘… complexity of internal models will allow companies to game the system …’
‘SST leads to complexity where simplicity is required …’
‘SST will increase the minimum Solvency level …’
Some have a somewhat reluctant attitude:
We would like to thank Andreas Kull (Ernst&Young) for the permission to use this slide
47
Impressions from the Industry
‘…facilitates more efficient use of risk capital …‘
‘Facilitates company wide risk culture and dialogue…’
‘… will reward companies that have a comprehensive risk management in place…’
‘… internal models are an excellent management tool and can be a competitive advantage…’
‚Rating dependent premiums will gain acceptance.‘
‘Increased transparency in the insurance sector may reduce cost of capital for the sector as a whole…’
‘… will lead to increased transparency in an insurer's financial strength/weakness…’
‘… is an effective regulatory instrument to prevent insolvencies…’
Some see it in a positive light:
We would like to thank Andreas Kull (Ernst&Young) for the permission to use this slide
48
Impressions from the Industry
„Wir haben diesen Sommer viel gelernt über unser Versicherungsgeschäft und über die Bedeutung von einzelnen Zahlen. Es gab viele Diskussionen über Kennziffern usw. welche zu einem Wissensaufbau in unserer Geschäftsführung führten und dazu beitragen werden, dass wir die Gesellschaft mit noch besseren Entscheidungsgrundlagen führen können. Die Ergebnisse aus dem SST-Testlauf nutzen wir auch für Diskussionen mit dem Verwaltungsrat (es gibt eine zusätzliche Sicht auf den Vermögensstand und den Geschäftsverlauf). Ich bin überzeugt, dass der SST die Führung von unserer Gesellschaft zukünftig unterstützen wird. Die Aufsicht lieferte uns dementsprechend ein weit ausgebautes Führungshilfsmittel.“ Comment by Martin Rastetter from the ‘Metzgerversicherung’ a small-
to-midsized nonlife company
Implementation of the SST for small to midsized companies:
Days of worked used approximately:
Internal: 40-50 days, External: 15-20 days
49
Results of Field Tests
• Market risk is often dominating (50%-80%)• The risk margin is between 10%-40% of target capital rsp. 1%-8% of
best-estimate provisions• Diversification Effect between Insurance and Market Risk: between -
5% and -30%• Effect of Scenarios on ES[RBC]: Between 5% and 50%
– Seems too high and some scenarios will be adjusted
• ‘Market consistent value of assets / Statutory value of assets’: between 90% and 120% but in most cases market consistent value is higher than statutory
• ‘Market consistent value of liabilities / Statutory value of liabilities’: between 70% and 100%
– While SCR/Target Capital requirement is in many cases higher than statutory capital requirement, the economics solvency ratio can be better or worse than the statutory solvency ratio (Solvency 1), since economic capital is in most cases substantially higher than statutory capital
50
Results of Field Tests: Target Capital
Credit Risk: In the standard model additiv add-on based on Basel II, but portfolio models can be used
Effect of scenarios
Expected technical &
financial result
Diversification between market and insurance risk
Insurance Risk, taking into account diversification between risk factors and lines of business and branches
Market risk, taking into account diversification between different asset classes Risk
Margin
1 year risk capital
Target capital
Relative length of bars correspond approximatively to actual median values
51
Results of Field Tests: Market Risk
Relative length of bars correspond approximatively to actual median values
Scenarios
Expected Result
Diversification between Market and
Insurance Risks
Insurance Risk
Market riskRisk
Margin
1 year risk capital
Target capital
Credit Risk
Interest Rates (CHF, EUR, USD, GBP)
Spreads
Real Estate
FX
Shares
Others
Diversification between market risk factors
Market risk
Diversification effect between different asset classes
52
0 0.2 0.4 0.6 0.8 1 1.2 1.40
0.2
0.4
0.6
0.8
1
1.2
1.4Results of Field Tests: Valuation
The following graph shows how market consistent liabilities compare to statutory liabilities.
In most cases, market consistent valuation releases substantial amounts of hidden reserves to risk bearing capital
Nonlife Best Estimate
Statutory Reserves (=100%)
Market Consistent Reserves (best estimate + risk margin)
Nonlife Life
Nonlife Risk Margin
Life Best Estimate
Life Risk Margin
53
0 0.2 0.4 0.6 0.8 1 1.2 1.40
0.2
0.4
0.6
0.8
1
1.2
1.4
Results of Field Tests: Valuation
The following graph shows how market consistent assets compare to statutory assets.
In most cases, market consistent valuation releases substantial amounts of hidden reserves to risk bearing capital
Nonlife Best Estimate
Statutory Assets (=100%)
Market Consistent Valued Assets
Nonlife Life
Nonlife Risk Margin
Life Best Estimate
Life Risk Margin
54
0 1 2 3 4 5 6 7 8-0.4
-0.2
0
0.2
0.4
0.6
0.8
1
Results of Field Tests: Diversification
Diversification between Market and Insurance Risk
Figure shows 1 year risk capital due to market and insurance risk (normalized with total 1 year capital requirement) and diversification.
Nonlife Market Estimate
Nonlife Insurance Risk
Life Market Risk
Life Insurance Risk
Diversification
Nonlife Life
0
100%
-30%
55
Results of Field Tests: Scenarios
Equity Drop -50%
Longevity
Global Deflation
US i.r. crisis 1992
Crash 2000/2001
Russia Crisis/LTCM 1998
European FX Crisis 1992
Nikkei Crash 1990
Crash 1987
Real Estate Crash 1987
Effect of Scenarios Expressed as Fraction of RBC
High ImpactNo Impact Medium Impact
56
Results of Field Tests: Sensitivities
Mort
alit
y
Longevit
y
Morb
idit
y
React
ivati
on
Lapse
Opti
on E
xerc
ise
Mort
alit
y
Longevit
y
Morb
idit
y
React
ivati
on
Lapse
Opti
on E
xerc
ise
Individual Business
Group Business
Sensitivities of life companies to different insurance risk factors, expressed as fraction of risk bearing capital (normalized with volatility of risk factors)
Example shows 4 different life companies
57
CHF EUR USD GBP Spread FX Shares Real Estate Others0
0.5
1Insurer 1
CHF EUR USD GBP Spread FX Shares Real Estate Others0
0.5
1Insurer 2
CHF EUR USD GBP Spread FX Shares Real Estate Others0
0.2
0.4Insurer 3
CHF EUR USD GBP Spread FX Shares Real Estate Others0
0.5
1Insurer 6
Results of Field Tests: Sensitivities
Relative influence of stand-alone market risk factors:
•Interest rates (CHF,EUR,USD,GBP)
•Spreads
•FX
•Shares
•Real Estate
•Others
Shown are standard deviations (in arbitrary units)
Correlations between above risk factors not taken into account
58
Contents
•Designing a Solvency System•Concept of the Swiss Solvency Test•The SST Standard Model•Experiences from the Field Tests• Internal Models
– Definition of an Internal Model– The Scientific Method– Acceptable and Unacceptable Models– Review of Internal Models
•Group Aspects
59
Internal Models
A model consists of :
• Methodology: Assumptions, models, mathematics, mapping of the real world to a conceptual framework,…
• Parameters: estimates, mortality tables, claim size estimates,…• Data: Position data, data on financial instruments, insurance
policies,… • Implementation: Software code, IT platforms, data warehouses,…• Processes: Testing, back-testing, falsification, plausibilisation,
estimation,…
A model is a framework to discuss economic capital
The point of the model is not (solely) the calculation of economic capital but to have a common framework for discussion of risks, of dependencies, of links between different areas of the business etc.
60
Internal Models: Problems
• How to ensure that the results are comparable between different companies
• How to ensure, that a company is not punished if it models risks more conscientiously than its peers
• How to be able to distinguish between acceptable and not acceptable models
• How to be certain that a model is deeply embedded within a company
When allowing internal models for target capital calculation, the problems a regulator faces are:
61
Acceptable Models: The Scientific Method
• Model verification is impossible
• Falsification is in many cases unpractical
• The scientific method can not be formalized. There can be no set of guidelines codifying the model approval process
There are limits on what a regulator can demand:
• We need to accept that some properties of a model can not be ‚proven‘ statistically (e.g. some dependency structures, some parameters)
• Models can, however, be persuasive
• Develop a set of model requirements any internal model needs to satisfy to be accepted by the regulator
• Supplement the model requirements by guidelines encouraging a process within a company so that models go through an appropriate internal review process (Pillar 2 and 3)
Therefore:
62
Internal Models
Even worse than having a bad model is having any kind of model – good or bad – and not understanding it
The review of internal models will focus strongly on how internal models are embedded within companies and how they are understood on different levels:
• Understanding of the underlying methodology, assumptions and limitations by the board, the CEO, the CRO, the actuaries etc.
• Quality of documentation, reports etc
• Use of the internal models by the company, e.g. for capital allocation, performance measurement etc.
63
Internal Models: Public Transparency
The public disclosure requirements on internal models should be principles based. The amount of information to be disclosed should be based on the principle that a knowledgeable person can get a reasonably good impression on the basic methodology of the internal models as well as on the major design decisions. In particular a description of the following main features should be provided:
• valuation methods (for assets and liabilities);
• risk measure;
• criteria for the choice of parameter values and distribution functions;
• major scenarios and risk factors and the assumptions on their dependencies;
• aggregation methods;
• embedding into the company's risk management processes;
• scope of the model and which relevant risks are not quantified.
64
Contents
•Designing a Solvency System•Concept of the Swiss Solvency Test•The SST Standard Model•Experiences from the Field Tests• Internal Models•Group Aspects
– Diversification– Modelling of Groups
65
Parent Company: Standard model can be calibrated using local, country specific statistics and models
Scope of Regulatory Models: Group
Branches: Can be in all parts of the world, home country regulator can not calibrate easily (if at all) a standard model to different risk profile. Mix of parent country risk to risks emanating from branches is widely varying from company to company
Subsidiaries: Can be in all parts of the world, home country regulator can not calibrate easily (if at all) a standard model to different risk profiles. Mix of legal entity risk to risks emanating from subsidiaries is widely varying from group to group. Capital flow between subsidiaries and parent is restricted.
Capital can flow (nearly) freely between branches and parent company and legal entity can be considered to be one risk-entity. Diversification between parent and branches.
Risk specific standard model is feasible
Risk specific standard model for legal entity is very difficult to develop
Risk specific standard model for group is extremely difficult to develop since in addition to legal entity model restrictions on fungibility of capital need to be taken into account
Parent CompanyLegal Entity
Group
Branch
Branch Branch
Branch
SubsidiarySubsidiary Subsidiary
66
Group Diversification
Legal Entity 2
Legal Entity 3Group Level DiversificationIntra-group risk and capital
transfer instruments allow group-level diversification to be realized and allocated to legal entities
Intra-Group Capital and Risk Transfer Instruments:
• Intra-group Retrocession
• Guarantees
• Participations
• Dividends
• Loans
• Issuance of surplus notes
A group is defined not only by its legal structure but also by its web of intra-group capital and risk transfer instruments
Legal Entity 1
• securitization of future cash flows / earnings
• sale / liquidation of a business
Limited fungibility of capital due to regulatory restrictions have to be considered
Capital and risk transfer instruments can only be considered if they are legally binding and acceptable to the regulators involved
Group
Risk reduction due to group effects
67
Group Diversification: Possible Approaches
“One group one balance sheet”
“Rational” Group “Darwinistic” Group
= SST Approach
The group is assumed to always support its legal entities even if no formal guarantees or obligations exist
The risk that a group will not support its legal entities or that a legal entity has to support the group is captured by regulators via group risk
The group is assumed to transfer capital only via legally enforceable capital and risk transfer instruments. Diversification is ‚transported‘ to and from legal entities only via formal, legally enforceable risk and capital transfer instruments
Regulator assumes that capital flows only via existing guarantees and obligations
Group risk is to a large part taken into consideration in assumptions
The group is assumed to support or let go of legal entities based on legally enforceable risk and capital transfer instruments and, where those not exist, according to some rational strategy, amenable to modeling.
Rational behavior has to be modeled using some utility function based on cost or benefit of legal entities going into run-off or being supported in case of financial problems
68
Group Diversification
Change due to risks ceded
Change due to contingent capital issued
Change due to contingent capital received
Change due to risks assumed
Allocated group diversification
Target capital net
Target capital gross
Target capital after allocation of group diversification
Within the SST, group diversification is the effect, legally enforceable group risk and capital transfer instruments have on required and available capital
69
Interesting Problems for Actuaries
Group issues:
• How to model the rational strategy of a group
• Optimizing structure of a group (branches/subsidiaries and web of risk and capital transfer instruments. How to set risk and capital transfer?
Internal Models:
• Improving on methodology of internal models; develop consistent firm-wide economic models
• Modelling the strategy and risk during a longer (>1 year) time horizon
• Allocation of capital
Other issues:
• Optimizing ALM using sophisticated financial products
• Integrating different fields (mathematics, finance, capital management, accounting, IT, risk management,…)
• Increased need to be able to communicate with different stakeholders (CEO, CRO,CFO, analysts, regulators, …)
70
For More Information
Philipp Keller: [email protected]+41 31 324 9341 / +41 76 488 3141
Thomas Luder: [email protected]+41 31 325 0168
Mark Stober: [email protected]+41 31 323 5419
Web-Links:www.bpv.admin.chwww.sav-ausbildung.ch (-> documents)
71
Notation
A(0) market value of assets at t=0
UPR unearned premium reserve (at t=0)
P estimate for premiums earned during year
K estimate for costs during year
RI Asset returns during year (r.v.)
DCY(1) discount factor at t=1 for the CY claims (r.v)
r1(0) risk free interest for one year duration at t=0
SCY claims during year (Current Year), r.v.
DPY(1) discount factor for PY-claims at t=1, r.v.
dPY(0) discount factor at t=0
RPY(0) best estimate of PY liabilities at t=0
CPY* RPY(0) re-evaluation of RPY
(0) at t=1 (r.v.),
( (1 - CPY) * RPY(0) = run-off result)