Casualty Actuarial Society - PowerPoint by pzl37647

VIEWS: 0 PAGES: 23

									PwC
  Casualty Actuarial Society

Practical discounting and risk adjustment issues
  relating to property/casualty claim liabilities

  Research conducted by PricewaterhouseCoopers
  Presented by Sam Gutterman, FCAS, FSA, MAAA    NCCI
                                      October 12, 2004
          IASB Board – February 15, 2005



                        PwC

                        PwC
          Outline

 • Project objectives & specifications
 • Measurement approaches
 • Findings
 • Significant issues




Slide 2
          Research objectives

     1. Given a set of fair value principles, evaluate the effect of
        discounting and risk adjustment on loss and loss adjustment
        expense (LAE) claim liabilities of US property/casualty
        insurance companies
     2. Identify significant issues associated with discounting and risk
        adjustment in loss and LAE claim liabilities




Slide 3
          Measurement objectives

  •       Active markets for claim liabilities do not exist
          − Entity-specific experience was used
  •       Claim liability measurement components
          1. Undiscounted estimate of future payments
             − Assumed undiscounted losses were estimated appropriately
          2. Discount for time value of money
          3. Margin for risk and uncertainty (“Market Value Margin” / “MVM”)
             − Did not reflect correlation across lines of business
  •       No adjustment for own credit risk other than in aggregate cost of
          capital


Slide 4
          Research specifications

          • Used publicly available entity-specific loss data only
             – Schedule P from US regulatory Annual Statements
          • Lines of business studied
             – Personal Auto Liability (shorter tail, although in some countries
               this coverage would be considered long tail)
             – Workers Compensation (long tail – stable)
             – Medical Malpractice, claims-made (long tail – volatile)
          • Evaluated ten companies for each line of business




Slide 5
          Measurement approaches

 Liability elements studied and measurement approaches evaluated
 • Discount factor models
    − Duration
    − Matched to yield curve
 • MVM models
    − Development (Mack) method using standard deviations
    − Stochastic simulation using standard deviations
    − Stochastic simulation using percentile distribution
    − Return on capital




Slide 6
            Measurement calibration

• Calibrated to the Cost of Capital Method at calendar year-end 2002
   – Capital equal to US regulatory minimum Risk Based Capital
     requirement
   – 10% target rate of return on capital for each company
• Arithmetic average of calibrations for 3 companies
   – 1 large, 1 medium, 1 small

                          Development Method   Stochastic Method   Stochastic Method
             Class            SD Multiple         SD Multiple          Percentile

Personal Auto Liability          1.2                  2.0                90%
Workers Compensation             1.0                  1.5                92%
Medical Malpractice              1.5                  2.3                95%



  Slide 7
                     Findings – discounting

                            Average discount factor by year for Personal Auto Liability


                    10.0%
                    9.0%
                    8.0%
  Discount factor




                    7.0%
                    6.0%                                                                  Duration
                    5.0%                                                                  Matched
                    4.0%
                    3.0%
                    2.0%
                    1.0%
                    0.0%
                              1998        1999        2000        2001        2002

                                                      Year
Slide 8
          Findings – discounting
 • Given a payment pattern, well-defined approaches are available
 • In general, no significant differences between duration and
   matching approaches were identified
 • Discounted results are affected by
      – Interest rate fluctuations
      – Shape of the yield curve
      – Expected payment pattern and tail (time until claim resolution)

 • Income affected by mixed attribute model

 • Significant uncertainty can be associated with payment patterns
      – In most cases, will be less than for amount of ultimate losses
      – Will require additional calculations, but generally not onerous
Slide 9
           MVM by company (increasing by size)
           Personal Auto Liability
                        MVM by Company at Year-End 2002
         100%
         90%
         80%
         70%
         60%
 MVM %




         50%                                                        ROC MVM
         40%
         30%
         20%
         10%
           0%
                0   2       4        6         8          10   12
                                   Company

Slide 10
           MVM by company (increasing by size)
           Workers Compensation
                       MVM by Company at Year-End 2002
        100%
        90%
        80%                                                    ROC MVM
        70%
        60%
MVM %




        50%
        40%
        30%
        20%
        10%
         0%
               0   2      4         6        8       10   12
                                 Company

Slide 11
              MVM by method for one company
              Personal Auto Liability
                        MVM by Method - Company C(M)
        18%

        16%

        14%
                                                              Develop SD
        12%
                                                              Stoch SD
MVM %




        10%                                                   Percentile
        8%                                                    ROC

        6%

        4%

        2%

        0%
               1998   1999      2000      2001         2002
                               Year-End

  Slide 12
           MVM by method for one company
           Workers Compensation
                     MVM by Method - Company K(L)
        70%

        60%
                                                       Develop SD
        50%
                                                       Stoch SD
MVM %




        40%                                            Percentile
        30%                                            ROC

        20%

        10%

        0%
              1998   1999     2000     2001     2002
                            Year-End

Slide 13
            MVM by company for one method
            Personal Auto Liability

                    MVM by Company at Each Year-End (Development Model)
                                                                            A (S)
          120%
                                                                            B (S)
          100%                                                              H (S)
                                                                            I (S)
          80%
MVM (%)




                                                                            V (S)
                                                                            C (M)
          60%
                                                                            F (M)
          40%                                                               G (M)
                                                                            D (L)
          20%
                                                                            E (L)

           0%
                                                                            Average
             1997        1998    1999     2000      2001    2002     2003
                                         Year-End

Slide 14
                MVM by company size
                Personal Auto Liability
                               MVM for All Companies and All years
          70%

          60%

          50%
                                                                                MVM vs.
                                                                                Company
MVM (%)




          40%
                                                                                Size
          30%

          20%

          10%

          0%
                0    200,000         400,000        600,000   800,000   1,000,000
                                         Reserves ($000)

 Slide 15
                 Discounting & MVM by company
                 Workers Compensation
                             FV Factor by Company at Year-End 2002
              80%


              60%


              40%
FV Factor %




              20%                                                          ROC FVF

               0%
                     0   2         4         6         8         10   12
              -20%


              -40%
                                          Company

  Slide 16
           Findings – MVM measurement

• Indications for MVMs varied, sometimes significantly
     − By method, for a given company and year-end
     − Over time, for a given company and MVM method
• The ranking of size of MVMs by method tended to vary over time
     − No method was consistently the highest nor the lowest
• For smaller companies, MVMs tended to be larger (when measured
  as a percentage of claim liabilities)




Slide 17
           Findings – effect on claims liabilities

• Personal Auto Liability
   – FV claim liabilities were generally greater than
     undiscounted/non-risk adjusted claim liabilities
• Workers Compensation
   – FV claim liabilities were generally less than or close to the
     undiscounted/non-risk adjusted claim liabilities
• Medical Malpractice claims-made
   – We did not consider the results of our testing to be meaningful as
     there was too much statistical variation in results
• Impact of moving to fair value of claim liabilities tended to be greater
  (due to larger MVMs) for smaller companies
                 Based on the model calibrations
Slide 18
           Findings – effect on incurred losses

• Current accident year incurred Fair Value losses were generally
  greater than undiscounted/non-risk adjusted losses
   – Relativities affected by calibration benchmark applied
• Accident year FV liability development benchmark would not be to
  zero
    – Due to relative changes in discount and MVM
• Leveraged effect of changes in claim liability would likely increase
  volatility of incurred losses




Slide 19
           Significant issues
           Modeling measurement

• Real data issues
• Measures of variation
   − Releasing constraint of acceptance of booked claim liabilities as
     expected (unbiased) ultimate amount of losses will affect
       Expected payment and claim notification patterns
            Variability of experience in relation to expectations
    − Variation from the tail/prior accident year bucket
           − Affected by study period, level of aggregation and degree of
             homogeneity of claims selected
    − Risk and variation inherent in certain claim liabilities are not be
      easy to analyze quantitatively (e.g. asbestos and environmental)

Slide 20
           Significant issues
           MVM estimation

• Variety of approaches exist, but no single approach currently
  preferred or accepted
• Further professional research, guidance and education needed
  regarding acceptable methods and calibration procedures for
  calculating MVMs to obtain consistent and comparable results
     – Single industry guideline for all lines of business and companies
       unlikely to be appropriate
• Calibration of MVM models
     – Challenging
     – Can significantly affect the results

Slide 21
           Significant issues
           Financial statement presentation
 • Based on currently available approaches, non-additivity of MVMs
   requires judgmental allocation among
    • Accident years
    • Lines of business
    • Business units
 • Accident year development disclosures may be confusing
    – Development of prior claim liabilities would not necessarily be
      benchmarked to zero
    – Component analysis of one-year development of prior year-end
      claim liabilities quite complicated
    – Solution might be triangular analysis shown on an undiscounted,
      non-risk adjusted basis

Slide 22
      Assessment of P&C actuarial methods

           Estimating undiscounted claim liabilities   Good


           Discounting estimated future payments       Good


           Estimating market value margins             Developing


           Calibration of MVM methods                  Emerging




Slide 23

								
To top