Our Changing Future ICA

1
Our Changing Future - ICA

• Peter McDade
• Neil Meldrum
• David Stevenson
• All views are those of the authors and do not
  necessarily represent the views of their employers

2
Agenda
David Stevenson

• Background to ICA
• Main challenges
• ICG Process
• Operational Risk
• Using the ICA
• Possible future developments
• What does it all mean for actuaries?

Peter McDade
Neil Meldrum
3
ICA and ICG

• The Individual Capital Assessment (ICA) is a
  firms own assessment of its capital
  requirements, given its risk exposures
• Self-assessment introduced with effect from
  31.12.2004 (GENPRU 2.1.6)
• Individual Capital Guidance (ICG) is any guidance
  provided by the FSA on the amount or nature of
  capital resources to be held by the firm under
  SUP 9.3 (eg. as a result of review of the ICA)
• ICG is not published (private between firm and
  FSA)

4
ICAS - Motivation

• To provide firms with an incentive to improve
  risk management
• Past weaknesses eg. provision for guarantees
• Limited engagement with operational risk
• Reinforce responsibililty of senior management to
  manage capital resources of the firm in line with
  its risk
• No longer acceptable to rely on compliance with
  regulatory minimum requirements
• Firms should consider their own risk exposures
  and form their own views on the amount and
  quality of capital they should hold
• To help inform FSAs own view of overall capital
  adequacy of firm
• Together with wider supervisory view, helps in
  providing Individual Capital Guidance

Sarah Wilson speech to ABI ICAS conference
06.03.2007 (paraphrased)
5
ICA

• Internal risk based capital assessment
• Considers all major risks explicitly
• Market
• Credit
• Insurance (mortality/longevity, morbidity,
  persistency, expense)
• Liquidity
• Group
• Operational
• Can allow for exercise of Management Actions
• Responsibility of Board (advised by Actuarial
  Function Holder)
• Responsibility to notify FSA if ICA has fallen,
  or is expected to fall, below ICG (SUP App 2.7)

6
FSA Regulations

• No detailed rules
• Three main ICAS principles high level guidance
  on their implementation (INSPRU 7.1)
• Supplemented by Guidance on ICAS produced by ABI
  other trade bodies
• BAS Guidance (GN46/GN47)
• Indication of FSAs approach to Principles Based
  Regulation ?

7
ICA Key challenges

• Absence of detailed rules
• Deciding what approach to follow
• Model calibration
• Data limitations
• Risk aggregation/correlations/non-linearity
• Subjectiveness/application of judgement
• Engaging senior management
• ICG Process

8
What approach to follow?

• Instantaneous stress, simulations of t1 balance
  sheet or run off?
• Measurement Value at Risk or Tail VaR ?
• Confidence level/time horizon
• Aggregation of risk (correlation matrix,
  copulas)
• ICG set using Value at Risk approach at 99.5
  confidence level over 1 year time horizon

9
What approach to follow?

• Most companies built on existing RBS or Economic
  Capital models
• Capital requirements for individual risks
  determined by applying instantaneous shocks to
  economic balance sheets (VaR approach)
• Analogous to RCM stress for RBS
• Shocks calibrated to be equivalent to a 0.5th
  percentile event over a one year time horizon
• Allowance made for management actions taken to
  mitigate impact
• Overall capital requirement calculated by
  applying correlation matrix to individual capital
  reqs
• Capital ??i,j KiKj 1/2

(?i,j are correlations between risk factors, Ki
capital reqs)
10
Build-up of ICA - illustration
Operational risk
Diversi- fication
Insurance risk
Mgt actions
Credit risk
ICA
Market risk
Longevity/Mortality Morbidity Persistency Expense
Liquidity risk Group Risk typically zero or
v.small
11
Choice of model

• Investment risks normal, lognormal, percentile
  from from ESG output, Jarrow-Lando-Turnbull for
  credit spreads,etc ...
• Is model tractable?
• How well does model fit historic data?
• Insurance risk
• Analysis of A/E may help in setting
  mis-estimation stress
• Trend risk more difficult (eg. future mortality
  improvements of Long cohort with floor,
  internal cause of death models)
• Operational risk new territory for many
  insurers

12
Data limitations

• Lack of data trying to calibrate a 1 in 200
  year event!!
• Relevance of data
• Most acute for non-investment risks
• Example - pandemic event (Avian flu H5N1)
• How relevant are past pandemics?
• 1918/19 Spanish Flu (40m deaths)
• 1957/58 Asian Flu (2m deaths)
• 1968/1969 Hong Kong Flu (1m deaths)
• What could 1 in 200 year event look like?
• Re-assortment of H5N1 genes could result in
  extremely virulent strain capable of transmission
  between humans
• Greater infection rates due to urbanisation,
  increase in travel
• Availability of antiviral drugs, medical advances
• Improved monitoring, transmission containment
  measures
• Different impact on different age groups?
• Need to apply judgement

13
Model calibration

• Even where data exists, how relevant is it?
• Example choice of time period for investment
  data
• Should periods like World Wars be excluded?
• Should we include periods of high UK inflation?
• Shorter period gt more volatile stresses more
  volatile ICA result, but more relevant to current
  market conditions
• Longer period gt more stable stresses ICA
  result, but possibly less relevance to current
  market conditions
• Question of philosophy/judgement

14
Inconsistency of data

• Example Correlation between Equity Returns and
  Fixed Interest Returns
• Positive for most of 20th Century
• Including during 1929 Wall Street Crash, 1973/74
  Middle East Oil Crisis, 1987 Black Monday,
  1997/1998 Russian debt/LTCM crisis, 1990s
  recession in Japan
• Negative in period following bursting of dotcom
  bubble in 2001/2002
• What is an appropriate stressed correlation
  assumption?

15
Risk aggregation/correlations

• Diversification benefit often very large item on
  ICA balance sheet
• Most companies use correlation matrix approach
• Judgement in setting stressed correlations
• Correlation matrix approach doesnt pick up
  non-linear interactions between risks
• Non-linearity could lead to understatement (or
  overstatement) of ICA
• Need to investigate make adjustments if
  appropriate

16
Non-linearity

• Scenario testing approach
• Combined market scenarios calibrated to 99.5/one
  year confidence level using results of individual
  market stresses
• Medium bang approach
• Calculates adjustment factor by comparing result
  of
• Running individual stresses at lower confidence
  level (eg. 93), and
• Combined scenario where all happen simultaneously
• Killer scenario
• Similar to medium bang but prob of stresses
  weighted towards most significant risk exposures

17
Engaging senior management

• Time constraints
• Initial education overhead
• Ongoing commitment to review exercise judgement
• Provides a common language for management of risk
  across business

18
ICG Process
Internal Planning
Submission Request
Initial Review
FSA Initial View
Meeting With FSA
Written Questions
Indication To Firm
FSA Panel
ICG issued
19
ICG Process How was it for you?

• More resource intensive than expected
• Open discussions
• Drip-feed of follow-up questions
• FSA views evolved during process
• Outcome no surprise

20
ICG

• First round now almost complete second round
  begun
• ICG provided to
• 53 life companies (95 of market by liabilities)
• 102 general insurers (97 of market by net
  premium income)
• Average ICG (ICG issued in 2006 Jan 2007)
• General insurers 110 of ICA
• Life insurers 114 of ICA (range 100 to 170
  a few outliers)

Sarah Wilson speech to ABI ICAS conference
06.03.2007
21
ICG

• Main reasons for ICG add-ons
• Life
• Aggregation (diversification, stressed
  correlations non-linearity)
• Operational risk
• General
• Operational risk
• Lack of evidence to support choice of assumptions

Sarah Wilson speech to ABI ICAS conference
06.03.2007
22
(No Transcript)
23
(No Transcript)
24
Asking the impossible?
25
Quantifying Operational Risk

• Op Risk risk of loss from inadequate or failed
  internal processes, people and systems, or from
  external events
• Who cares?
• Could adopt simple factor-based approach (eg x
  of assets)
• Motivation for more advanced approach
• Internal
• Better understanding of risk
• Effectiveness of controls
• Incentivise risk management as controls lead to
  lower capital
• External
• FSA
• Emerging best practice
• Ratings agencies?

26
Cant get there from here?

• How to get from this
• ...to this?

Capital requirement x Assets
99.5
0.5
Capital requirement
27
Main stages of the journey

• Design a new process
• Identify the key risks
• Collect data (actual loss data expert
  testimony)
• Create loss distributions
• Build use a Monte Carlo model
• Identify required capital at 99.5th percentile

28
Data the key issue

• Data is scanty even for market risks
• Much less available for OR
• Internal loss data inadequate by definition
• External loss data useful, but not enough
• Expert judgement subjective but vital to plug
  gaps
• Use everything available, but biggest challenge
  is to harness the expert knowledge in the
  company.

29
Key risk identification

• Filtering process trying to identify the most
  significant risks
• Focus on loss events not strategic risks
• For risk management purposes the whole chain of
  causality is important.
• For the modelling we focused on the immediate
  prior cause (eg poor complaints handling)
• This is where the financial loss actually occurs
• More concrete and therefore easier for people to
  come up with numbers
• Identified collection of front line risks eg
• Business interruption
• Poor complaints handling
• Legal risk
• Fraud
• Pandemic Flu

Failure to set strategic direction
Low morale
High staff turnover
Staff shortage
Poor complaints handling
30
Scenario workshops

• Series of workshops set up involving risk owners
  / experts in each field
• Tasked with creating loss distributions for each
  of the front line risks
• Facilitation challenge strike balance between
• Motivation (this affects actual capital perhaps
  SP rating)
• Terror (this affects actual capital/rating!!)
• Asking the impossible?
• at least 3 points on a loss distribution for each
  risk most likely, severe, extreme events
• need both impact and probability for each point
• use experience of actual losses existing
  controls
• brainstorming, no wrong answers, what if
  analysis
• knowledge, enthusiasm, creativity

31
Example of a loss distribution
32
Monte Carlo modelling - best way to combine the
loss distributions

• Inputs to model
• Loss distributions
• Correlation assumptions
• Each simulation produces an aggregate loss across
  all risks
• Run 10,000 simulations
• Now have aggregate loss distribution
• Capital requirement 99.5th percentile loss

Note can allow for correlations between risks
33

• Useful to have drill down facility can examine
  component loss events for any simulation
• Management challenge - is this scenario plausible?

34
Embedding

• Such a process should become a core part of risk
  management
• Rapid development in recent months will want to
  go round loop again this year to ensure robust
• Management review of data modelling results
• Challenge / feedback / refinements
• Not an exact science a means to an end
• Improving understanding of the key risk drivers
• Help to formulate articulate risk appetite
• Connecting up operational and financial areas of
  the business
• Eg value of good insurance policies gt no changes
  should be made without considering capital
  implications.
• Improved collaboration across functions.
• Helping you to understand
• for some risks holding capital is unavoidable
• for others better solution is strengthen controls
• tradeoffs between the cost of improving controls
  and the savings in capital cf cost/benefit of
  hedging market risk

35
Are we there yet?

• Journey will continue
• Can always improve on the modelling, but the data
  is a bigger challenge
• Not impossible!
• Scope for creativity

36
Final Destination Reached?
37
Using the ICA (Embedding)

• ICA is more than just a number, its about good
  risk management.
• ICA is superior to the old Pillar 1
  calculations as it measures the capital required
  based on the risks inherent in a companys
  business.
• But the measurement of risk is only one part of a
  strong risk management structure.
• The first wave of reviews largely focused on the
  measurement of risk

38
Using the ICA (Embedding)

• . However, the second wave of reviews will be
  different.
• we are looking at how best to add real value
  to the next round of reviews of ICAs...scope
  for added value from taking a more qualitative
  approach challenging firms on how they are
  using their ICA in practice to make better
  business decisions and improve risk and general
  governance.
• Sarah Wilson, Director Retail Firms Division FSA
  (6th March 2007)
• So the bar is being raised by the FSA.

39
What have companies done so far?

• Strategic Decisions
• Review of risk appetite
• Information gleaned from ICA used to develop
  hedging strategies.
• Revised reinsurance arrangements.
• De-risking asset mix.
• Operational decisions
• Using in capital projection plans.
• Some companies have sought to embed economic
  capital into pricing.
• Used in the determination of investment policy

40
Regulations and Guidance

• FSA launched its new ICA Principles during 2006.
• Objective was to give greater clarity over FSAs
  expectations and help deliver consistent capital
  guidance across the industry.
• 3 sub-principles (rules), which can be summarised
  as
• Assessment must reflect the firms actual risk
  profile.
• Comparability to a 99.5 / 1 year probability
  that the value of the firms assets will exceed
  the value of their liabilities.
• Model methodology documenting the firms
  reasoning and judgement underlying the ICA
  assessment.
• The ABI also launched its A Guide to the ICA
  Process for Insurers, which aimed to provide
  advice to companies on how to interpret guidance.

41
Possible Future Developments

• ICA is still in its relative infancy and emerging
  best practice will continue to develop.
• Methodology and assumption changes are becoming
  less of a priority
• Short term challenges include
• (Refining) Projections of the ICA
• Developing robust analysis of change
• Enhancing operational risk methodology
• Developing scenarios
• More attention is planned to be given by the
  industry to improving integration and use of the
  ICA in the business

42
Possible Future Developments

• Future developments are likely to be influenced
  by Solvency II, as UK Insurers are likely to want
  to go down the internal model route.
• To have internal model approved for solvency II,
  there are three tests to be passed. They are
• Statistical Quality Test
• Calibration Test
• Use Test
• FSA to suggested to date that the hurdle to be
  passed for these is much higher than the UK
  industry has had to achieve so far under the ICAS
  regime.

43
Final Destination Reached?

• Still work to be done in embedding ICA by the
  industry
• Work required by industry to get internal model
  approval for Solvency II
• So the journey not yet complete.

44
What does this mean for actuaries?

• ICA is principles based not rules based.
• Requires a lot more judgement which have to be
  justified to Board (ultimate ownership).
• Quantifying the risk requires detailed modelling,
  but also understanding of the weaknesses of the
  model.
• Need knowledge of all the business
• Cannot be a one department number. Need to
  listen to the business to ensure that the risks
  are considered.
• Need to explain what the number means in business
  terms how does this number relate to risk in
  the business.
• Therefore, actuaries are very well placed for
  ICA/Solvency II with skills and knowledge we can
  bring to the table.
• This provides interest work and enhances
  marketability.
• Opportunities in Europe as they address Solvency
  II ?
• However, no room for complacency other skilled
  professionals in the risk assessment field.

45
Question Time.