Accounting for NonLife Insurance in Australia

1
Accounting for (Non-Life) Insurance in Australia
IASB MEETING, APRIL 2005OBSERVER NOTE (Agenda
Item 3a)

• Tony Coleman Andries Terblanche
• Presentation to IASB
• 19 April 2005

2
Agenda

• Australian Non-Life Insurance Business
• - Users of Discounting and Risk Margins
• Background to Australian Insurance Market
• Insurance Basics Why use Risk Margins ?
• Consistency Reliability
• Use of Accounting Results by Management
• Practical Issues
• In Conclusion The Vital Role of Disclosure

3
A Birds Eye View of Australian Non-life Insurance

• Relatively Small, but Sophisticated Market
• 11th largest insurance market in the world with
  direct, gross private sector premiums of approx
  A26Bn (US20Bn) (Australian Prudential
  Regulatory Authority Sept 2004)
• Enters the world top 10 after allowance for
  public sector (A5Bn of compulsory workers
  compensation and motor bodily injury insurance),
  Foreign-based business (Lloyds etc)
• Australia 1 of world GDP, 2 of global
  non-life insurance premiums
• Profitable (since 2001 !)
• Approx A3Bn of insurance profit in 2003
  (Insurance Council of Australia - Aug 2004).
  A4Bn in 2004 ?
• Industry underwriting profits in each of last
  three years
• First time for two consecutive years
  underwriting profits in industry in more than 25
  years
• A Wide Range of Products
• Personal Lines/Commercial Lines almost 50/50
• Short tail/Long tail around 60/40 by premium
  (including public sector etc)
• All standard insurance products sold anywhere in
  the world are freely available in Australia

4
A Birds Eye View .. Recent Developments

• Market Rationalisation
• Significant recent takeover/merger activity. The
  Big 5 now write approx 70 of the gross direct
  market premium (10 years ago the top 5 insurers
  wrote around 40)
• Four of the Big 5 are now Australian-owned (and
  publicly listed), increasing the focus on
  profitability. (10 years ago the majority of the
  top 10 insurers were foreign-owned)
• HIH
• The collapse of Australias 2nd largest non-life
  insurer in 2001 had significant ramifications for
  the market (availability of cover, pricing of
  products, public confidence in general)
• HIH was the first significant non-life insurance
  failure for over 20 years
• Lack of discipline in reserving (and hence
  pricing) practices was seen to be a significant
  contributory factor
• Misuse of finite (financial) reinsurance was a
  contributing factor
• Subsequent Royal Commission had many (61)
  recommendations

5
A Birds Eye View .. Current Issues

• Improved Prudential Regulation and Risk
  Management
• Not all as a result of HIH. Regulatory reform was
  already happening (although significantly
  enhanced by HIH fallout). APRA introduced major
  prudential reform from 1 July 2002.
• Increased audit intensity at a global level
  (Enron, Parmalat etc), reviews by ASX, the
  competition regulator (ACCC) and reform of
  selling practices (FSRA)
• Liability Tort Reform
• A reaction to the increasing compensation
  culture has seen legal restrictions introduced
  to control access to compensation for liability
  (casualty) insurance claims (especially relating
  to bodily injury.
• Each state has dealt separately with the issues
  leading to complicated effects on insurance
  pricing and a market expectation of reduced
  premiums
• Latent Claims
• Asbestos-related claims are creating a
  significantly increasing liability for past
  exposures (although not at USA levels)
• Potential future, as yet unrecognised latent
  claims are a growing concern for the industry

6
Accounting Standard (AASB 1023 before 1 Jan 2005)

• Assets at market value, through PL A/c
• Discounting of liabilities at riskfree interest
  rates (via PS300)
• Vulnerable to manipulation via financial (finite)
  reinsurance (as demonstrated by HIH)
• Silent on use of Risk Margins
• Market practice on risk margins varied widely
  HIH had none, but others maintained various PoA
  up to 90
• Australian adoption of IFRS from 1 January 2005
  and HIH fallout lead to a new AASB1023,
  requiring use of risk margins, disclosure of
  central estimates and liability Probability of
  Adequacy (PoA) disclosure requirements

7
A Reminder of Insurance Basics

• Outcomes of risks from individual policies are
  unknown when underwritten
• However, when many similar risks are
  underwritten, expected results of total portfolio
  become more predictable
• Claims are driven by
• - Frequency (or probability) of a claim event
  occurring and
• - Severity (or size) of a claim if it occurs
• Risks inherent in different classes of insurance
  vary over a spectrum
• - High frequency / low severity (eg motor)
  outcomes relatively easy to predict reliably
• - Low frequency / high severity (eg earthquake)
  outcomes harder to predict reliably

8
Simple Illustration of the Insurance Risk Process

• To help discuss the concept of accounting for
  uncertainty
• Assume we roll a dice 100 times to represent the
  results of underwriting 100 insurance policies in
  one year.
• If 1 is result, insurer pays a claim of 1
• If 2 is result, insurer pays a claim of 2
• If 3 is result, insurer pays a claim of 3
• If 4 is result, insurer pays a claim of 4
• If 5 is result, insurer pays a claim of 5, BUT
• If 6 is result, there is no claim at all.
• Assume all claims will be paid 1 year after
  policies are underwritten and
• That the dice can be rolled at any time during
  that year.

9
Illustration

• What is the liability of the Insurer after all
  policies are written but BEFORE any dice have
  been rolled ?
• The higher the amount reserved the greater the
  probability will be that there are adequate
  (sufficient) funds to pay all claims.

10
Illustration Level of Reserving (before any
dice are thrown)

• From the well known nature of this distribution
  we can confidently predict that results will be
  such that
• Probability of Adequacy
• 50 250
• 75 262
• 90 272

11
Illustration The Law of Large Numbers
Reliability of Estimates CoVs

• Coefficient of Variation (CoV) used to measure
  Volatility
• C o V Standard Deviation of Distribution
  / Mean of Distribution
• Note that Standard Deviation Square Root
  of Variance
• If only 1 throw Mean 123450 / 6
  2.5
• 2 2 2 2
  2 2
• Variance (1.5)(0.5)(0.5)(1.5)(2.5)(2.5)
  / 6 2.917
• Hence CoV (Square Root of 2.917) / 2.5
• 1.707 / 2.5
• 0.68 Hence CoV 68 of the
  mean

12
Illustration The Law of Large Numbers
Reliability of Estimates - CoVs

• If we have 100 throws
• Mean 100 x 2.5 250
• Variance 100 x 2.917 291.67
• Standard Deviation Square Root of 291.67
  17.08
• C o V 17.08 / 250 0.068 Here CoV
  6.8 of the Mean
• If we have 10,000 throws
• Mean 10,000 x 2.5 25,000
• Variance 10,000 x 2.917 29,167
• Standard Deviation Square Root of 29,167
  170.8
• C o V 170.8 / 25,000 0.0068 Now CoV
  lt 1 of the Mean

13
Illustration - Premium Profit

• For simplicity, assume all expenses of operation
  are incurred at outset and total 40.
• Suppose insurer charges 3 per throw to
  policyholders.
• Hence total premium 300
• Expected underwriting profit 300 - 250 -
  40 10
• (A lower profit will occur 50 of the time and
  higher profit will occur 50 of the time)
• Is 10 the profit that can be reported as earned
  ?
• If so, when can it be reported as earned ?

14
Capital Requirement

• Irrespective of the liability adopted for general
  purpose financial reporting, the insurer needs
  500 to operate without any probability of going
  bankrupt and has to meet expenses of 40 at the
  outset after receiving 300 in premiums paid at
  the outset.
• Hence, the actual Capital the insurer needs at
  the outset to be sure of being able to meet its
  commitments is
• 500 40 - 300 240
• This capital requirement is in addition to the
  cash generated by the contracts at the outset of
  
• 300 - 40 260

15
Projected Profit / (Loss) Return on Capital

• Assume no taxes are payable and that capital and
  cash flows can be invested at a risk free rate of
  5 p.a.
• Then the expected results will be as set out
  below
• Source of Profit Funds Interest
  Best Worst Expected
• Held Earned Profit/ Profit/
  Profit/ (_at_ 5) (Loss) (Loss) (Loss)
• Policy Contracts 260 13 273 (227)
  23
• Capital 240 12 12 12 12
• Total 500 25 285 (215) 35
• The expected rate of return on capital employed
  will be
• 35 / 240 14.6 p.a.

16
Level of Liability (before any dice are thrown)

• Assume that the insurer decides to adopt a
  liability at the outset equal to the central
  estimate of the liability of 250, plus a risk
  margin equal to its expected profit from the
  product contract cash flows of 23, all
  discounted at the risk free rate of 5
• Hence liability at the outset would be
• ( 250 23 ) / 1.05 260

17
Profit / (Loss) at Outset

• Hence, if the expenses of 40 incurred at the
  outset are expensed in full at the outset the
  profit reported at outset will be
• 300 260 40 Nil
• Alternatively, if the non-claim expenses are
  amortised over the year the profit at outset will
  be
• 300 - 260 40
• (Note however that this second treatment implies
  booking all premiums and
• full liability for claims expenses at outset,
  but deferring non-claim expenses
• which seems inconsistent)

18
Consistency of Liability PoA

• Setting the liability at the end of the year at
  273 by including a 23 risk margin in the
  liability equates to setting the probability of
  adequacy (PoA) of the liability marginally in
  excess of 90 (recalling that the 90 PoA was
  272).
• Hence the PoA can be used to consistently
  calibrate the size of risk margins for
  outstanding claims (where, by definition, there
  is no unexpired risk premium remaining in the
  cash flows).
• Further, this would be consistent with a fair
  value of the liability if the markets/insurers
  required rate of return on capital employed was
  14.6 p.a.

19
Illustration Summary of Insurance Liability
Issues

• We have shown how the uncertainties of the
  insurance risk process can be accounted for BUT
  we have only allowed for changes in our modelled
  outcomes assuming no change in the underlying
  process or its parameters. If the process
  parameters start to change, but the process is
  modelled correctly, this leads to Parameter
  Risk.
• We also need to account for the fact that our
  model of the system is likely to be less than
  perfect (e.g. perhaps the dice, unknown to us,
  was actually a loaded dice). This is Model
  Risk.
• Additionally, we have assumed in our example that
  the process does not change, but it probably will
  over time (e.g. a player moves address and
  forgets to role the dice (or lapses). This is
  Process Risk.
• In a true insurance risk situation, we need to
  account for all three types of risk in addition
  to basic random statistical fluctuations in
  outcome.

20
Critical Issues in Reserving

• Estimation of claims trends (frequency
  severity)
• Interest Rate used to calculate present values
• Measures of Uncertainty
• Probability of Sufficiency (50, 75, 90)
• Co-efficient of Variation (Std Dev/Mean)
• Market Benchmarks

21
Linking PoA With Market Value Why 75 PoA ?

• Compared
• NPV (at Cost of Capital) of requirement to hold
  funds in excess of the central estimate up to
  99.5th percentile of outstanding claims at Risk
  Free Rate
• i.e. initial capital needed less NPV of expected
  releases as claims are settled
• Assumed
• Claims log-normally distributed (i.e. skewed
  outcomes)
• CoVs constant over run-off of claims
• Realistic returns on capital
• Tested
• Short, Medium and Long Tailed classes

22
Linking PoA With Market Value - Conclusions

• For classes modelled
• Percentile liability representing a realistic
  result varies significantly by duration
• Around 55-60 for short tail
• Around 80-90 for long tail
• Around 75 is reasonable for a typical mixed
  portfolio with allowance for diversification
• 75 PoA is just one possible benchmark
• Risk Margins can be set in other ways

23
Balancing Liabilities and Assets

• Accounting for assets is easy just let the
  market (for the majority of assets) be the model
  (and the means of valuation)
• Even though there is no real market for most
  insurance liabilities we can measure the
  liability on a basis consistent with the asset
  approach
• Adequate Disclosure is the key !
• Discounting the liabilities at an appropriate
  rate (the risk free rate) is essential. Future
  cash flows can, and should, be derived for even
  the most simple of liability estimation methods,
  and hence enable the application of discount
  rates to all claim liabilities.
• Risk margins create the market value adjustment
  for the level of uncertainty adopted
• The value of non-life insurance liabilities
  should not change when backed by different types
  of assets.
• Probability of Adequacy is one way of providing
  the disciplined structure. There are others (e.g.
  Tail VAR or use of profit margins).

24
Using the Accounting Figures to Manage the
Business

• Results Tend to be Volatile
• But, so is the business!
• Discount rates may change with market movements,
  but active asset/liability management can
  ameliorate this effect
• A prospective approach to unexpired risk speeds
  up recognition of both profits losses
• Disclosure and Discipline (Actuarial Standards)
  are Vital
• -- Risk margins should not vary
  much over time as a percentage of the central
  estimate
• Transparency of reporting means that trends in
  business outcomes are recognised at an earlier
  stage (and therefore tend to have a lesser
  once-off effect on results)
• Diversification effects are still subject to a
  range of approaches
• Hence Result Smoothing is Difficult
• Everything is auditable!
• Internal and external reporting are entirely
  consistent

25
Consistency of Market Results

• Initial Regulators Data on Use of Risk Margins
  Shows Some Variation (Source APRA May 2003)
•  

26
Consistency of Market Results in Australia

• On a Practical Level, Consistency is Improving
• Central estimate is now clearly the mean of the
  distribution of potential outcomes (not the
  median, or the mode!)
• Before APRA there was a wide range of usage of
  risk margins. Now, 75 PoA for all insurance
  liabilities (including premium liabilities) tends
  to be the minimum.
• Companies are converging around 90 PoS of
  liability for outstanding claims as a market
  standard
• Market analysts (users) very interested in
  liability disclosures
• Diversification allowances are still an area of
  some divergence, but increased disclosure is
  starting to supply some answers
• Not uniformly transparent (yet), but getting
  there !

27
Practical Experience Central Estimates

• Few challenges to existing methodology in
  Australia
• Central estimate value has been adjusted for
  inflation and discounted at a risk free rate
  for many years
• Increased focus reinsurance to turn gross into
  net values
• Confirmed the central estimate as the mean of the
  distribution of potential outcomes
• Benefits
• Expanded actuarial influence to virtually 100 of
  insurance liabilities
• Improved internal management discipline
• Better quality and quantity of data
• Stronger communication links with Board and
  senior management
• More transparency of approach
• Developed linkage to risk margin work

28
Practical Experience Risk Margins

• Challenges have been significant and fundamental
• What does 75 (or 85 or 90) probability of
  adequacy mean and how should it be estimated? (We
  may be able to estimate parameter risk
  accurately, but what about model risk and process
  risk?)
• There was (alarmingly) little global literature
  upon which to base an approach
• The Institute of Actuaries of Australia and APRA
  contributed to research
• Adequate Actuarial Standards (APRA GPS210 and
  IAAust PS 300) were vital
• Diversification allowances are still a challenge
• Benefits have emerged
• Improved understanding of companies approaches
  (if not yet full consistency)
• Established thinking that a margin is needed
• More management focus on true drivers of business
  risk

29
Practical Experience Premium Liabilities(for
APRA reporting LAT for AASB 1023)

• Challenges were again fundamental
• A new discipline was needed (as actuaries had
  tended not to venture into the accounting
  preserve of unearned premiums)
• The methods were basically an extension of claim
  reserving methodology (but with some variation to
  allow for risk differences)
• The full challenge has yet to be met (since AASB
  1023 accounting standard is only just starting to
  examine risk prospectively)
• Benefits are already apparent
• Clear linkage with the pricing process
• Clarifies gross and net issues
• Encourages usage of dynamic financial analysis
  techniques (especially for low frequency claims)

30
Minimum Disclosure Recommended

• Central Estimate of Liability for Outstanding
  Claims
• Total Liability for Outstanding Claims
• (so that (1.) (2.) Total Risk Margin)
• 3. Estimated Probability of Adequacy
  (Sufficiency) of (2.)
• 4. Separate analysis of (2.) and (3.) shown for
  (a) total short tail
  (lt1 year to paid claim) portfolios, and
  (b) total long tail (gt 1
  year to paid claim) portfolios
• 5. Separate analysis of (1.) to (4.) by currency
  of liability
• 6. Details for each liability of mean term and
  discount rates and inflation rates (including
  judicial or superimposed inflation) used
• 7. Details of movement in central estimates over
  past year

31
An Australian Example of Disclosure Promina Ltd
- 31 December 2004
32
Promina Limited - 31 December 2004The
following average inflation (normal and
superimposed) rates and discount rates were used
in the measurement of net outstanding claims
33
In Conclusion

• Insurance Isnt Black and White
• Outcomes are nearly always uncertain
• If we introduce appropriate standards, we can
  account for the Grey
• By introducing the concept of a distribution of
  potential outcomes and requiring clear disclosure
  about PoS used
• Risk Margins provide the framework to achieve
  this
• Meaningful, bench-markable disclosure is key to
  consistency
• Management of the insurance business is improved
  by the added transparency and discipline
  introduced