IASB Educational Session Non-Life Claims Liability

Transcription

1 IASB Educational Session Non-Life Claims Liability Presented by the January 19, 2005 Sam Gutterman and Martin White

2 Agenda Background The claims process Components of claims liability and basic approach The loss development triangle Types of loss reserve methods Examples Some problems Disclosure Page 2

3 Background Scope Non-life insurance - Focus of this presentation is on general (property & casualty or P&C ) insurance Motor, personal property, fire, liability against suit, professional malpractice Liabilities for non-life insurance 1. Contract liability 2. Claims liability Focus of this presentation For other insurance, analysis is similar but uses simpler methods Page 3

4 Background P&C coverages Contracts are normally one year in duration - A few exceptions e.g., Japanese claim free savings contracts - Range from short-tail property to long-tail liability - Can include right of termination by one or both parties during the period Approaches described here also applicable to self-insurance and reinsurance Page 4

5 Background Claims groupings Similar types of claims are grouped to enhance accuracy of estimation process - By coverage With similar payment patterns If loss ratio is used, similar profit and expense margins - By period of claim incurral - But experience should be large enough to be reliable A single contract can contain several coverages - Automobile damage to vehicle, liability, collision, medical payments, uninsured motorists, - Commercial packages property and liability Page 5

6 Background Data organization Claims organized by - Coverage - Accounting (loss) date - Report date - Recorded date - Measurement dates Typical problems - Changes in products, terms/conditions, business mix, rates, people Page 6

7 Background Loss date Claims liability arises as of a loss date within coverage period Definition can vary by coverage - Date of occurrence (accident date) - Report date for claims-made contract - Manifestation date for certain sicknesses - Purchase date for retroactive contract Page 7

8 Background Loss year Types of claims groupings - Accident / loss / report year - Calendar period of loss - Underwriting / policy year - Contracts written in period - Overlaps two calendar years - Used by Lloyd s In some cases, quarterly or semi-annual periods Subsequently measured by period since loss date Page 8

9 The claims process 20-Dec-2003 Accident occurs 3-Jan-2004 Potential claim reported to insurer 6-Jan-2004 Insurer records claim information e.g. 2,000 Formula-reserve 19-Jan-2004 Claims adjuster estimates ultimate amount 10,000 case reserve IBNR 18-Jan-2005 Claim closed (Possible reopen later) 2-Jan-2005 Payment sent 0 case reserve 23-Dec-2004 Settlement agreed 25,000 case reserve Change IBNER 5-July-2004 Claims adjuster reestimates if new information emerges Reinsurer notified if involved 20,000 reserve Change IBNER Page 9

10 Components of a claims liability Claims liability (often referred to as loss reserve) 1. Estimated ultimate payments, less payments to date or 2. Case reserve (assigned by adjuster or formula) + Bulk or IBNR reserve, theoretically separate but usually combined Incurred But Not Enough Reported (IBNER) - Case reserve development Incurred But Not Reported (or Recorded) reserve Includes associated expenses, possibly split - Allocable to claim (ALAE) - Not allocable to claim (ULAE) Page 10

11 Basic approach Actuarially sound reserves Based on estimates Derived from reasonable assumptions Using appropriate methods Inherent uncertainty In part due to time to settle Conceptually a range of values True value known only after all claims settled Page 11

12 Basic approach (continued) Estimates - Of ultimate losses by period of claims, their components, their timing, associated premiums (if loss ratios used) and their probability distribution (if stochastic methods used) - Data used almost always entity-specific where relevant - May be supplemented by industry experience when company doesn t have sufficient relevant experience - Limited relevant market experience - Based on financial reporting context, e.g., level of prudence Same approaches used for direct insurance, reinsurance and self-insurance Page 12

13 Basic approach (continued) Estimation principles are fairly simple - Group together similar coverages and claims - Evaluate how cohorts develop through historical patterns - Identify expected changes in internal and external factors - Apply past pattern to groups of outstanding claim groups - Result is the estimated liability The devil is in the details - Identify groupings properly - Determine relevant patterns - Reflect expected changes - Estimate future claim payments Page 13

14 Basic approach (continued) Claim emergence and loss settlement patterns vary significantly - Coverage contract terms and conditions - Direct or reinsurance assumed business - Underwriting intensity and mix of risks - Time period - Location country, state or neighborhood - Claims management and case estimate methodology - Mix of source of loss - Percent of asbestos claims caused by mesothelioma, or of complete property losses Impractical to separately quantify the effect of all of these Page 14

15 The loss development triangle What is it? Follows a cohort of similar claims incurred within a period through each subsequent reporting period to measure loss component patterns and trends to estimate patterns of future activity Can be used with - Cumulative paid losses - Incurred losses = cumulative paid losses + case reserves at end of each period - Number of claims - Size of claims - Case reserves Page 15

16 The loss development triangle (continued) Why is it used so often? - A useful way of both thinking about and presenting the emergence of claims information over time - Components usually exhibit identifiable trends over time - Makes estimating ultimate losses easier Number of years included in triangle varies by - Data available - Type of loss (e.g., speed of payment) - Relevancy of losses of older periods Diagonals represent status or activity at a measurement date or during specific calendar period Page 16

17 The loss development triangle (continued) Assumes claims experience exhibits patterns - Usually by period since claim date - Sometimes with calendar effect Change in claims management, e.g., settlement strategies or case reserve estimation techniques Inflation general or social Court / legal decisions Page 17

18 The loss development triangle An example Cumulative Paid (000) EZ INSURANCE COMPANY AUTO LIABILITY Accident Development State in Months Year Ultimate ? ? ? ? ? ? ? Page 18

19 The loss development triangle (continued) Patterns shown are multiplicative, evidenced by - Current period value divided by that of prior period - Link or loss development factors Ratios of losses at successive evaluations for a defined group of claims (e.g., accident year) The triangle is actually a rectangle - Experience in upper left triangle - Objective estimating bottom right triangle - Looks easy, but judgment always needed Page 19

20 Types of loss reserve methods 1. Case reserves - Based on claim-specific information available - Used in more mature periods 2. Loss ratio - Reserve = Ultimate loss paid through valuation date Ultimate loss = premiums x expected loss ratio - Sources include loss trend, rate change, industry, judgment, pricing assumption Page 20

21 Types of loss reserve methods (continued) 3. Development (chain-ladder) - Possibly the most common - Uses loss development triangles, with simple multiplicative link ratios - Applies to paid losses or incurred losses (cumulative paid + current case reserves) 4. Claim count x average size claim (frequency / severity) - Separately estimates number of claims and their severity - Can estimate each independently Uses loss development triangles - Ultimate loss = Ultimate count x Ultimate average size Page 21

22 Types of loss reserve methods (continued) 5. Bornhuetter-Ferguson - Weighted average (based on percentage expected to be reported) of Incurred (or paid) claims to date Initial (or updated) estimate of ultimate losses for claim cohort based on - Loss ratio or exposure x estimated loss per exposure - Used in most recent year or two or where limited reported losses to date such as assumed reinsurance 6. Many other methods Almost always multiple methods evaluated Page 22

23 Development method -- example Cumulative Paid (CU 000s) EZ INSURANCE COMPANY AUTO LIABILITY Accident Development State in Months Year Accident Paid Development Factors between months Year Ult Unweighted average point average Average excluding high/low Volume weighted average Selected LDFs Cumulative LDFs Page 23

24 Development method loss development factors Major assumption development is a multiplicative process Selected development factor for 2003, 12 to 24 month = 280 / 128 Cumulative development factor - Multiply all development factors from current point to last experience point times tail factor (to take to ultimate value) = x x Applied to relevant diagonal value Warning not all triangles this well-behaved Page 24

25 Development method -- example (continued) Reserves Based on Paid Development Method EZ INSURANCE COMPANY AUTO LIABILITY (CU 000s) Accident Paid Selected Estimated Unpaid Case Year to Date CLFactor Ultimate Losses Reserves (1) (2) (3) (4) (5) (6) from the diagonal cumulative LDFs (2) x (3) (4) - (2) , , , ,342 1, Total 3,436 7,302 3,866 1,770 Page 25

26 The loss development triangle The tail Tail factor takes losses from experience in the triangle to their ultimate value - In example, takes losses after 7 years to their ultimate value - Difficult with limited data - Use industry or similar entity experience if relevant Results can be quite sensitive to tail - Because applied to cumulative losses - For example, a 2% change could result in a 40% increase in IBNR How Much Tail Can There Be? Development in Reinsured Layers Selected Cumulative Age to Ultimate Factors Source: Reinsurance Association of America data Line of Business 15 Years to Ultimate 25 Years to Ultimate Workers Compensation General Liability AL Treaty Page 26

27 Expected loss ratio method Expected loss ratio equals - Expected ultimate losses / Earned premium, or - 100% - contract expense ratio tax ratio expected profit margin IBNR = Expected ultimate losses paid losses case reserves Basis - Pricing assumptions, trended claim and costs, industry data - Since treatment of contract liability can influence earnings of premium, contract liability can influence loss ratio Useful if - New or radical change in coverage - Immature accident years for long-tailed coverages Page 27

28 Expected Loss Ratio Method example Earned Premium = 100,000 Expected Loss Ratio = 65% Paid Losses to date = 10,000 Case Reserves = 13,000 Total Reserve = (100,000 x 0.65) - 10,000 = 65,000-10,000 = 55,000 IBNR Reserve = 55,000-13,000 = 42,000 Page 28

29 Bornhuetter-Ferguson method Weighted average of results of development method and loss ratio method - Weight based on expected percentage reported at valuation time for a given accident period - Uses historical experience directly to estimate ultimate losses Several alternatives to loss ratio value can be used - Exposure x Exposure units (such as contracts inforce) - Expected claim frequency x expected average size claim - Prior estimate Used when insufficiently reliable reported claim experience to date or long-tail business, but don t want to ignore current results Page 29

30 Baseball example How many home runs will Barry Bonds hit this year, given: - He has hit 20 home runs through 40 games - There are 160 games in the season Information needed for estimate based on various actuarial methods - Expected Ultimate Value = 40 (before season started) - Factor to project Current Diagonal to Ultimate Value Current Diagonal = 20 home runs hit through ¼ of season Development method: 80 = 20 x 4 (4 = cumulative development factor) Loss ratio method: 40 (initial expectation, unadjusted by experience) Bornhuetter-Ferguson: 50 = 20 + ¾ x 40 (weighted average of experience and initial expectation, times percent of expected unreported) Choice depends on knowledge of his current health, whether he hits better in the spring or in the fall, Page 30

31 Some problems Choice or use of methods No method is perfect and no method can be applied in all situations In practice, common to use different methods for different periods since accident year, e.g. - Case reserves for oldest years - Development method for years Bornhuetter-Ferguson for most recent year Examine results from several methods for each time period Significant effect of changes in estimates on reserves and even greater effect on income Page 31

32 Some problems Uncertainty Significant uncertainty in estimates - Although a single number is needed for balance sheet - A range may be more appropriate, can be based on - Different methods, LDFs, probability distributions, - Judgment is always needed Must understand the business, claims process and environment If something can go wrong, it will Common to continue to refine data, assumptions and methods Page 32

33 Some problems Uncertainty and stochastic methods With increasing emphasis on effect of risk, stochastic methods have become more common Requires estimates of probability distribution of claim liability components - Ultimate loss, frequency, severity and timing Results in estimates of frequency distribution of possible outcomes - Can produce confidence intervals, with enhanced understanding of risks But danger of overstating the precision of the resulting estimate based on a further set of assumptions Page 33

34 Key Assumptions Potential Problems Claims settlement patterns unchanged Case reserving practices & philosophies unchanged No claim processing changes Policy limits have no impact on loss development Increased delays in claim closing rates. Conscious effort to improve case reserve adequacy Introduction of new case reserving procedures. Unreported change in data processing Increase in claim backlog Revised claim payment recording procedures. Changed mix of full policy limits claims or changed policy limits. Page 34

35 Key Assumptions Potential Problems Loss development unaffected by changing loss cost trends No change in mix of business Surges in inflation Increased litigation or new court rulings Diminished policy defenses. Changed reinsurance program Increased long-tail exposures Introduction of new or revised coverages. No cyclical loss development Underwriting cycles impact pricing adequacy, claims settlement or reserving practices. Page 35

36 Key Assumptions Potential Problems No inflation or if there has been, fully reflected in experience to date Inflation general or social/legal, that varies by calendar period or unequally by coverage types No data anomalies Catastrophic or unusual losses reflected in loss experience Unusual claim settlement/reporting delays Unusually large claim(s) If reinsurer, reporting problems with direct writers or lack of sufficient data Page 36

37 Monitoring results Retrospective testing Important to continue asking how assumptions performed - Through continual challenge of your own assumptions Applicable to each claim liability component Examples of questions to answer - How many new claims were expected this year - What average claim size was expected - Why was there a deviation from estimates Due to random fluctuation or changes in conditions that need to be reflected next time? Page 37

38 Disclosure The best way to monitor loss development for most lines is through loss triangles - Power and importance of how did we estimate - Retrospective loss triangles can show changes in loss reserve estimates over time - Meaningful period typically five or ten years, depending on the speed of claims closure, but earlier period may no longer be relevant Although discounted reserves can be used in a triangle, care is needed to compare diagonals on an apples-to-apples basis Page 38