Current Loss Reserving Developments CAS Annual Meeting November 15, 2005 Chuck Emma, Pinnacle Tom Ryan, Milliman John J. Kollar, ISO ISO Study of Loss and Loss Adjustment Expense Reserves A. Industry Schedule P (Net) B. Analysis of Direct Losses CAS Annual Meeting November 15, 2005 John J. Kollar, ISO A. Industry Loss Reserve Analysis • • • • More than 900 insurer groups Year-ended 12/31/04 Schedule P data compiled by A. M. Best More than 95% of LLAE reserves for studied lines Lines studied • • • • PP Auto Liability HO/Farmowners Com. Auto Liability Other & Products Liab. Claims-Made • Other Liab. Occurrence • Com. Multi-Peril • • • • Med Mal Occurrence Med Mal Claims-Made Products Occurrence Reinsurance (NonProportional Liability) • Workers Compensation Some Key Points • Excludes reserves for environmental and asbestos (E&A) claims – Possibly $30B to $50B deficient • Analysis assumes incurred losses from 9/11 were fully developed at year-end 2004 – Estimated direct insured losses: $20B to $30B – U.S. net insured losses: $6B to $9B • Adjustments have been made for other major catastrophes Methodology • Paid link-ratio technique • Case-incurred link-ratio technique • Consistent with ISO study of 2003 data Factors affecting analysis • • • • Data quality Development factors Tail factors Professional judgment Conventions • Each deficiency/redundancy expressed as percentage of indicated undiscounted reserve as estimated by ISO – Positive percentages indicate deficiencies – Negative percentages indicate redundancies Summary Preliminary Indications of Reserve Deficiencies Paid • Lines Studied • All Other Lines • Total – all lines + 2% + 3% + 2% • In Dollars $9B – (excluding E & A ) Case Incurred + 7% + 3% + 7% $31B Perspective • Reserve adequacy has improved for 3 consecutive years – Reserves were about 3 percentage points more adequate at year-end 2004 than at year-end 2003 – Reserves were about 11 percentage points more adequate at year-end 2004 than at year-end 2001 Preliminary Indications by Line • Lines with deficiencies Paid • • • • Products Occurrence Com. Multi-Peril Workers Comp Reinsurance (Non-Prop.) + 7% + 1% + 9% +25% Case Inc. +12% + 4% +15% +33% Preliminary Indications by Line • Other Lines • • • • • • • Priv. Pass. Auto Liability Homeowners/Farmowners Commercial Auto Liability Other Liability Occurrence Claims Made Other & Prod. Medical Malpractice – Occ.* Medical Malpractice – C-M.* Paid Case Inc. - 4% -16% -10% - 3% - 8% - 3% -11% - 6% -10% - 1% + 5% + 1% +12% - 9% *ISO still investigating whether reserve adequacy is overstated consequent to data anomalies. LALAE Ratios: Accident Year vs. Calendar Year % 100 95 90 85 80 75 70 65 60 Reserve adequacy deteriorated for at least 6 years but then improved in 2002, 2003 & 2004. 1996 1997 1998 1999 Accident Year 2000 2001 2002 2003 2004 Calendar Year Loss Reserve Changes vs. Industry Profitability, All Lines % Changes in reserves are correlated with profitability. 35 30 25 20 15 10 5 0 -5 '71 '74 '77 '80 '83 '86 '89 '92 '95 '98 Change in LLAE Reserves / Paid LLAE GAAP Return on Average Net Worth '01 '04 Retrospective Estimated Deficiencies & Economic Discount, All Studied Lines % Discounted reserves were inadequate from 2000 to 2003. 30 25 20 15 10 5 0 -5 -10 1996 1997 1998 1999 2000 Paid Link Ratio Compound Discount Factor 2001 2002 2003 2004 Case-Incurred Link Ratio B. Analysis of Direct Losses • Segment Analysis – State/coverage/class group/etc. • Reserving/Benchmarking – Comparable mix of business • Tail Factors • Confidence Intervals Schedule P Lines with ISO Distributions • Homeowners/Farmowners - Homeowners • Private Passenger Auto Liability/Medical • Commercial Auto/Truck Liability/Medical • Commercial Multiple Peril - Commercial Multiple Peril Liability - Commercial Multiple Peril Property Schedule P Lines with ISO Distributions (Cont’d) • Medical Malpractice - Occurrence - Hospitals - Physicians - Surgeons • Other Liability - Occurrence • Products Liability - Occurrence • Auto Physical Damage - Commercial Auto Physical Damage - Private Passenger Auto Physical Damage Schedule P Lines with ISO Distributions (Cont’d) • Special Property - Fire - Allied Lines - Inland Marine Segment Analysis – Link Ratios 14 Period 1 To Ultimate Chain Ladder Link Ratios 12 Ratio 10 8 6 4 2 0 Sch. P Oth. Liab. Occ. ISO Prem Ops All Tables W. CRR ISO Prem Ops Table 1 ISO Prem Ops Table 2 ISO Prem Ops Table 3 ISO Prem Ops Tables 1, 2, & 3 The Schedule P data is net, includes Composite Rated Risks (CRR), and is evaluated as of 12, 24, etc. months. The ISO data is direct, excludes CRR (except as noted), and is evaluated as of 15, 27, etc. months. Reserving/Benchmarking • Used aggregate direct data by segment – State/coverage/class/etc. – Paid/incurred/losses/claim counts/LAE • Weighted aggregate direct data by each insurer’s unique mix of business – Losses rather than link ratios – Separately for each accident year • Applied link ratios based on aggregate data to an insurer’s data to calculate an indicated reserve Triangles, Link Ratios, Averages Triangle of developing losses Link Ratio Factors Link Ratio Averages Benchmarking of Reserves Tail Factors • Modified Bondy Method The ultimate factor (UF) determined using the first prior factor (FF) and the second prior factor (SF) as follows: If SF > 1 and [ 0.8 * LN(SF) >= LN(FF) >= 0 ] or SF < 1 and [ 0.8 * LN(SF) <= LN(FF) <= 0 ] then UF = FF ^ { LN(FF) / [ LN(SF) – LN(FF) ] } Otherwise, UF = FF ^ 4 • Development beyond 10 years – up to 20 years using direct data Development 10 vs. 20 Years Relative Volatility of Tail Factors Development of Parameters for Confidence Intervals Estimated parameters for claim severity and frequency separately • • • • By line By settlement lag (valuation) Industry for severity By insurer size for frequency Reserve Risk: Average size and volatility of open claims increases over time Claim Amount Big Claims Settle Slowly 95th % Mean 0 1 2 3 Open After n Years 4 5 6 Development of an Insurer’s Confidence Intervals • An insurer’s loss and loss adjustment expense reserves - By line - By accident year (latest valuation) • Reinsurance arrangements - Retention - Coinsurance - Per claim limit • Scale factors to reflect differences in average severity Measures of Variability Frequency/Severity/Reserves Measure of variability in aggregate reserves Measure of variability in frequencies Measure of variability in severities Measure of variability in reserves by line and period Confidence Interval and Aggregate Loss Reserve Distribution Cumulative Probability Aggregate Loss Reserve Distribution 1.0 0.9 0.8 Expected Loss 0.7 0.6 0.5 0.4 50% Confidence Interval 0.3 0.2 0.1 0.0 1,000 1,100 1,200 1,300 1,400 1,500 Loss Reserves ($Millions) 1,600 1,700 1,800 Confidence Intervals for Loss Reserves – Effect of Reinsurance Intervals for Small Products Liability Writer 45,000,000 Without Reinsurance With Reinsurance 35,000,000 25,000,000 15,000,000 5,000,000 70% 75% 80% 85% 90% 70% 75% Confidence Intervals 80% 85% 90% Confidence Intervals for Loss Reserves – Effect of Size Confidence Intervals for General Liability* (Relative to the Aggregate Mean) 1.8 Smaller Insurer 1.6 Larger Insurer 1.4 1.2 1.0 0.8 Aggregate Mean = 604,383,911 0.6 Aggregate Mean = 40,292,261 (15 times the smaller insurer's volume) 0.4 70% 75% 80% 85% 90% 70% 75% Confidence Intervals * A mix of Other Liability and Products Liability. 80% 85% 90% Applications • Reserving triangles – Combine segments using each insurer’s unique mix of business • Individual segment analysis • Tail factors • Confidence intervals (ranges around expected) • Benchmarking for CEO/CFO, Board, investors, rating agencies, regulators, etc. Future Plans • Explore ways of improving reserve estimates – Credibility weighing an insurer’s data with larger data sets – Generation of more refined confidence intervals – Additional tail factor treatments – Correlations between lines/dependencies • What else would be valuable for loss reserving?