Actuarial Judgment What Is The Financial Impact?

Actuarial Judgment What Is The Financial Impact? Mujtaba Datoo, ACAS, MAAA, FCA

Why Apply Judgment? • Predicting unknown events or outcomes • Incomplete information

10

Actuarial Judgment

Common Uses in Actuarial Analyses • Estimate unpaid liabilities – Current vs. Long Term Liability – Discounting

• • • • • •

Establish future loss funding Experience modification Modeling risks Capital (surplus) targets Retention (SIR) analysis Collateral analysis, Loss Portfolio Transfers 11

Some Key Areas of

Actuarial Judgment • • • • •

Loss development Trend Credibility Determining data appropriateness Selection of models – Confidence levels – Increased Limits Factors

• Surplus targets

12

Developing Losses to Ultimate Selection Uses Actuarial Judgment • Which method most appropriate? – E.g. paid or incurred loss development – What averages to use – long-term, recent – Ultimate losses should converge

• How much weight to give each method – Check for ranges – Low, middle, high, “central estimate"

• Impact on funding, liabilities, rates

13

Loss Development Curves WC, GL, Auto, Med Mal

14

Trend • Calculates changes in cost over time – E.g. CPI, labor costs (for auto damages) • Many tools exist – Regression analysis • What data is most appropriate if entity’s own data is sparse – Similar entities? Statewide? • Is the trend line going to behave as predicted? – Bent line, increasing, decreasing • Factor in external information – Economic conditions, management input (layoffs?), frozen payroll, new exposures, etc.

15

Trend, graphically

actual

10% reform impact, moderating trend

projection

16

Credibility • Weight assigned to predictability of given body of data • Several statistical approaches to measure this – Requires judgment for contextual usage • E.g. Experience modification

• Does the weight comport with objective – E.g. Stability of rates

17

Experience Modification

Laden With Judgmental Selections • Typical formula: (Member’s experience) x (credibility weight%) + (“some proxy”) x (100% less member’s weight)

• All of these involve some judgment: – How many years of experience? – Cap member’s losses? – How much weight to give member’s claims experience? – How much weight to give member’s exposure? – Which proxy to use for “credibility complement?”

18

Data Anomalies Treatment • Do you smooth large one-off claims? – How do you treat catastrophic claims?

• Do you develop these claims or rely on expert claims adjusters or counsel’s expertise? • Do you exclude and develop and trend the rest and then add back in? • Each treatment approach will have impact on final result! – Are results reasonable? Material?

19

Model Selection • Models based on selection of parameters • Statistical tests to see if it fits data • Sensitivity testing of outcomes based on various parameters • Are results consistent and meet reality check

20

Modeling Basics • • • •

Use data that reflects entity’s experience and objective Fit data to “theoretical” distribution, e.g. Pareto Split losses into frequency and severity Select parameters for frequency and severity – Empirically – Statistical distributions, e.g. Poisson and Lognormal

• Simulate and test results

21

Curve Fitting, Histogram of Claims 7,000 claims from last 10 years 2003 to 2012 1% of claims > $100K, and costs 33% of incurred $$

70 claims > $100,000 & incurred = $16M out of 48M

22

Confidence Level Expected Value (average)

Probability

50th Percentile Median

Loss

50%

70%

90% Factor of 1.10 Factor of 1.38

23

Confidence Levels • Percentile matching: – “expected”, 70%, 80%, 90%, etc.

• Based on claims experience or model results • Check $$ difference between selected confidence level and “central estimate” – Does the margin meet risk tolerance? – Does the margin meet risk exposure?

24

Pricing Legislative Changes • Examples where data and judgment is used to estimate impact of legislative/judicial changes: – WC benefit reform – Tort thresholds increase, e.g. Florida $100/200 to $200/300 effective 2011 – Med Mal tort – limiting impact of non-economic changes

• Data may not be available under new law – Estimate based on related experience – Simulate scenarios

25

How to Test if Judgment is Reasonable • Results expected in a likely outcome or range – Based on a priori information – Law of large numbers • Results will gravitate within a narrow range

– Knowledge of similar programs – Test results against emerging experience

• Management expectation of program behavior

26

Check Results of Actuarial Judgment • What is the upper or lower bound of factors selected judgmentally? • Are the results material? • Does emerging experience comport with assumptions?

27

Where Does Judgment Come From? • Training • Experience – Selected factors have to borne with reality of similar programs

• Professional standards of practice • Expert input – Research materials, studies – Claim specialists, legal counsel

28

10,000 Hour Rule A common theme that appears throughout Outliers is the "10,000-Hour Rule", based on a study by Anders Ericsson. Gladwell claims that greatness requires enormous time, using the source of The Beatles' musical talents and Gates' computer savvy as examples.

29

10,000 Hour Rule – The Beatles The Beatles performed live in Hamburg, Germany over 1,200 times from 1960 to 1964, amassing more than 10,000 hours of playing time, therefore meeting the 10,000-Hour Rule. Gladwell observes that the time The Beatles spent performing shaped their talent. He quotes Beatles' biographer Philip Norman saying "So by the time they returned to England from Hamburg, Germany, 'they sounded like no one else. It was the making of them.'"

30

10,000 Hour Rule - Gates Gates met the 10,000-Hour Rule when he gained access to a high school computer in 1968 at the age of 13, and spent 10,000 hours programming on it.

31

End Objective • Meet your financial obligations in light of uncertainty – Provide “best” estimates that meets objectives – Not to reduce costs from the estimation process • Other mechanisms to reduce costs: loss control, risk transfer, collateral negotiation, etc.

• Others rely on these estimates – Auditors, rating agencies, reinsurers, regulators, stakeholders

• Communicate assumptions, changes, impacts – 2-way communication: you to the actuary and actuary to you!

32

Questions?

Mujtaba Datoo

Actuarial Practice Leader (949) 608-6332 [email protected]

33

Thank You!

34