EXAM 5 - Final Consolidated Set

Rates of trend example calculations Common assumptions for trend

"0.5% quarterly change trended for 3 years" --> (1 + 0.005) ^ (3*4) "2% continuously compounded change trended for 3 years" --> e^(.02*3) Common Assumptions: 1) Policies written uniformly over time 2) Premium earned uniformly and losses occur uniformly over the policy period

How are groups of policies dealt with when calculating premium/exposure metrics?

"15th of the month rule" / "24th's method" - treat all policies as if they were written on the mid-point of their respective period Ex: monthly basis - assume everything written on 15th of the month --> only a good assumption if policies written evenly throughout the period

Accident Year Aggregation

(AKA calendar-accident or fiscal-accident year) Premium and exposures the same as Calendar Year; loss transaction only for accidents that occurred during the twelve month period, regardless of when policy was issued or claim was reported *AKA calendar-accident year or fiscal-accident year *better match of premium and losses than calendar year *development is necessary (not fixed at year end) *good for isolating major events like catastrophes

Policy Year Aggregation

(AKA underwriting year when looking at the date when a reinsurance policy became effective) Considers all premium and loss transactions on policies written during a 12 month period, regardless of when the claim occurred or when it was reported, reserved, or paid. *premiums and exposures fixed at year end, but losses continue to change after policy year *BEST match between losses and premiums (a true match) *takes longest to develop: premium and exposures not fully earned until 24 MONTHS after the start of the policy year --> development subject to more uncertainty than AY (shortcoming) & more immature when thinking about development

Earned exposures facts (same for premium)

*assumes probability of claim distributed evenly throughout the year --> doesn't hold for warranty lines or lines with seasonal fluctuations in writing (ex: boat owners insurance) *all PY earned exposure is assigned to the policy year that the policy was written, whereas CY earned exposure can be earned in two different years --> DON'T BOOK NEGATIVE EARNED PREMIUM FOR CANCELLATIONS

Why might it be useful to use commercial lines rating mechanisms in addition to or instead of a manual rate?

*exposure to loss varies significantly from risk to risk *classification plans aren't as refined due to limited data *individual insureds are large enough such that their own experience has some credibility in predicting future experience

What are some things to consider when thinking about trends?

*premium trends are simpler and thus more stable than loss trends *will past trends continue into the future? *how much history should be used? *BALANCE STABILITY WITH RESPONSIVENESS

Difference between Calendar Accident Year & Accident Year

*premium/exposures are developed for accident year (ex: audits) *normally used interchangeably

Objectives when aggregating data for ratemaking purposes

1) Accurately match premium and losses for the policy 2) Use the most recent data available (i.e. can we use recent data or do we have to develop it) 3) Minimize the cost of data collection and retrieval

What methods are used to aggregate exposures and premiums?

1) Calendar year (SAME FOR EVERYTHING EXPOSURES & PREMIUM as accident/calendar-accident year) and 2) policy year Report year is a loss concept ONLY

How do we adjust for shock losses in ratemaking?

1) Cap losses at basic limits (minimum amount of insurance that's legally allowed) then separately derive rates to price for losses above basic limits (increased limit factor pricing) ***IMPORTANT: also adjust historical premium to basic limits rates when calculating loss ratios*** *this method doesn't work for WC as there are no policy limits ***to get basic premium ex: suppose we have premium for a policy with $5000 limit and basic limit is $1000 --> divide out $5000 limit rating factor and multiply by $1000 limit rating factor to get the adjusted premium 2) Cap losses at some determined amount & apply an excess loading to cover expected losses above the cap level 3) Remove ground-up shock losses then add back in a loading for them FOR #2 & #3: use actuarial judgement, maybe look at percentile of size of loss distribution and consider percentile above which losses become volatile

Compare the methods for on-leveling premium

1) Extension of Exposures *re-rate all historical policies at the INDIVIDUAL level, recalculating premium to the amount that would have been charged under current rates *more accurate *requires detailed data & making assumptions for new rating variables w/ no historical data --> may be difficult to incorporate changes in schedule rating guidelines (commercial lines) IDEA: the on-level factor is just the ratio of premium at CRL to premium at historical levels for a given year 2) Parallelogram Method *adjusting historical premium by an average factor for each period on GROUPED data *quicker and easier calculations *might not be appropriate for Class level ratemaking (ex: Territory A different from Territory B) --> rate changes might have varied by class *assumes policies written evenly across periods --> bad assumption for seasonal lines

What are the possible pricing methods (least to most popular)?

1) Guessing (if no relevant data is available) 2) Non-insurance data 3) Competitor rates (not great bc there are differences in UW guidelines, expenses, mix of business, etc.) 4) Industry data (NCCI, ISO, etc.) 5) Use the insurer's historical loss ratios or pure premiums adjusted to reflect the future period

How do we adjust for CAT losses in ratemaking?

1) Non-modeled CAT losses: more frequent events like hailstorms are dealt with similarly to shock losses, and exposure growth in catastrophe-prone areas must be considered 2) Catastrophe models: estimate losses from rare events (earthquakes, hurricanes, etc.) where even long-term averages don't have enough data --> run simulations on insurer's book of business *no such thing as an "average" CAT year: either a little or a LOT *non-pricing ways to mitigate risk: reinsurance, raising deductibles, restricting writing in risky areas

Criteria for exposure base

1) Proportional to expected loss Ex: 2 house vs. 1 house better than $200,000 house vs. $100,000 house 2) Practical - objective, easy to obtain and verify Ex: car-years better than estimated annual miles because people could lie 3) Considers preexisting exposure base established within the industry Ex: don't want large premiums swings for individual insureds, expensive IT costs, complications for future analysis

What are the main principles of P&C ratemaking (last bullet: another point to consider when balancing fundamental insurance equation)?

1) Ratemaking is prospective (restate historical experience and project expected value of future costs, not trying to recoup past losses) 2) A rate provides for all costs associated with the transfer of risk (losses, claims, UW expenses, profit, etc.) 3) A rate provides for the costs associated with an individual risk transfer (ideal but not always possible, sometimes need to work in groups) 4) A rate is reasonable and not excessive, inadequate, or unfairly discriminatory if it is an actuarially sound estimate of the expected value of all future costs associated with an individual risk transfer --> Balance should be attained at the individual and aggregate levels (equitable rates, don't want high and low risk groups balancing each other out)

Non-pricing actions used by insurance companies to minimize potential catastrophe impact

1) Restrict the writing of any new business 2) Require higher deductibles for catastrophe-related losses 3) Purchase reinsurance. In addition to these non-pricing actions, the actuary may alter the underwriting profit provision in the rates to reflect the higher cost of capital needed to support the extraordinary risk caused by the higher concentration of policies.

What are anomalies in the context of loss data? What is our goal when considering them?

1) Shock losses: large loss from individual claims *definition varies by insurer and LOB --> $1mm loss means less to $1b book of business than $100m book *ex: permanent disability of young worker in WC 2) Catastrophe losses: large losses from many claims *defined by ISO Goal in pricing: produce rates that cover the costs of shock losses/CAT over a long period of time i.e. don't overestimate/underestimate future losses when these events do/do not occur

Which factors do premium development depend on?

1) The type of plan permitted by the jurisdiction or offered by the carrier 2) The stability of the historical relationship between the original premium estimate and the final audited premium 3) Internal company operations (auditing procedures, marketing strategy, accounting policy, etc.)

What are the most common trend scenarios (from period and to period)?

1) Trend earned premiums from historical CY to future PY 2) Trend losses from historical AY to future PY 3) Trend losses from historical PY to future PY (long-tailed lines) NOTE: all 3 are trending to FUTURE PY

What are the steps involved in handling a claim?

1. Assign claims to appropriate claims professionals/adjusters/examiners (usually specialized; might hire Independent Adjustor if understaffed) 2. Gather & document information about the claim and the policy 3. Determine whether the type of claim is covered by the policy 4. Quantify the damages covered by the policy and possible establish case reserve (ground-up then apply deductibles, coinsurance, limits, etc.) 5. Settle the claim 6. Collect any appropriate recoveries

What are the four steps in estimating ultimates?

1. Exploratory Data Analysis - identify key characteristics & anomalies and balance to verified sources 2. Apply appropriate techniques to estimate ultimates 3. Evaluate the conflicting results of the different techniques - reconcile and explain different outcomes 4. Monitor projections of actual vs. expected development - update or correct projections with new information

Which ratios are used as primary measure of the adequacy of rates?

1. Loss Ratio Losses / Premium Pure Premium / Average Premium *usually reported losses and total earned premium *LAE may or may not include losses 2. LAE Ratio LAE / Losses *benchmark for claim settlement procedures

Discuss some of the different claims handling philosophies

1. Staffing levels - tradeoff between salary expense and quality of handling 2. Case reserving: conservative vs. aggressive 3. Customer experience: some insurers may be more lenient in paying claims 4. Litigation strategy --> ex: emphasize making out-of-court settlements 5. Reliance on software in quantifying damages

Reserving dates & terminology

Accounting date: defines group of claims being analyzed & paid/unpaid split Valuation date: date through which transactions are included in the analysis Review date: actuary's work date (after the valuation date) reflecting material information included in the analysis *development = changes in values between valuation dates for a given accounting date *ex: actuary has data through 3/31/2014 (valuation date) to estimate unpaid claims as of 12/31/2013 (accounting date) from all accident years for which the insurer has liability _________________________ Estimated claims reserve: estimate resulting form an analysis Carried (booked) claims reserve: reserve value appearing on an insurer's financial statement

How do we deal with continuous changes and in what order? What are some examples of continuous changes? What kinds of data can we use for trending?

Adjusting historical data to reflect future expected mix of business and levels of social and economic inflation *do this AFTER adjusting for one-time changes, anomalies, and seasonality *use average premiums or pure premiums (not something like total premium that would be impacted simply by writing more policies) *inflation impact on payroll in WC *customers shifting towards higher deductibles *rising gas prices --> lower auto freq. *increased medical costs --> higher liability severity Historical insurer data, industry data, or economic data

Calendar Year Aggregation

All premium and loss transactions occurring during 12 month period without regard to date of policy issuance, accident date, or report date *everything fixed at year end and available quickly, no development involved *mismatches timing between premium and losses --> more useful for short-tailed lines

What is retrospective rating?

An insured's loss experience during the current policy term is used to determine their current policy premium *initial premium is collected at the start of the policy term *adjustments to premium based on reported loss experience (i.e. development occurs, change in case reserves, etc.) are made starting around 6 months after the policy expires and every 12 months thereafter as losses develop (applied through insurer refunds or extra payments by insured)

What is the basic goal of ratemaking? What should an actuary consider when determining the granularity of desired data?

Balance the fundamental insurance equation: Premium = Losses + LAE + UW Expenses + UW Profit Want homogeneous, credible groups of data that balance the fundamental insurance equation at both the aggregate and individual level

Why is the diagonal policy year line straight in our usual diagrams?

Because we assume that premium/exposures are earned evenly throughout the policy term (pro-rata)

What is on-leveling?

Bringing premium to current by restating is at current rate levels, accounting for the effects of recent historic one-time changes Usually done for earned premium 2 Methods: 1) Extension of Exposures 2) Parallelogram Method

Order the four data aggregation methods from least to most development

CY - no development RY - some development AY - more development PY - most development (more than PY b/c avg. accident date is later - takes two years to close)

4 loss data aggregation methods: paid and reported losses

CY: paid losses - all payments made during the calendar year on any claims; reported losses - paid losses plus change in case reserves during the year AY: paid losses - sum of all payments made on claims that occurred during the year; reported losses - paid losses plus case reserves as of the valuation date for those claims that occurred during the year PY: paid - sum of all payment made on policies that were written during the year; reported - paid + case reserves as of the valuation date for those claims that were written during the year RY: same as AY except "reported" ****NOTE THAT FOR ALL FOUR OF THESE YOU SUBTRACT OUT SALVAGE AND SUBROGATION

How do we adjust losses for benefit changes?

Can on-level reported losses for benefit changes (similar to how we on-level premium in the parallelogram method) *assume losses uniformly distributed over time period --> this assumption is normally only valid for accident quarters or policy quarters (accident quarters represented with straight lines, policy w/ diagonal?) Two ways benefit changes can apply 1) All losses on policies written after certain date --> diagonal line 2) All new losses occurring after certain date --> vertical line ***IF LOSSES ARE GIVEN AT PRE-BENEFIT CHANGE LEVELS, APPLY FULL CHANGE TO ALL YEARS --> DON'T NEED TO DO PARALLELOGRAM METHOD***

What are some ways to address an imbalanced fundamental insurance equation

Change premium *preferred way: simple and less consequences *change rates or restate exposures Change losses *reduce coverage or change UW guidelines Change LAE *more or less approach to defending claims and incurring legal fees, settlements, etc. Change UW profit target *maybe take a longer-term view on profit Change UW expenses *change commission rates or lay off employees Do nothing! (if profit targets are met and rates are competitive)

What are one-time changes and what are some examples?

Changes implemented or adopted by an insurer at a specific date in time that impact premium, loss, and/or exposures Rate changes *ex: changes to rating algorithm *in-force policies won't be recalculated until policy renews Law changes *stipulate change in coverage, benefits (ex: WC indemnity benefits), or rates --> law change to rates: implemented like a rate change or effective immediately for in-force policies --> law change to coverage or benefits: can affect all policies written after implementation date or be implemented immediately Court rulings *similar to law changes Expense changes *ex: company changes commission rate paid to agents

Claims definitions

Claim is a demand by the claimant (insured or 3rd party) for payment under the terms of the insurance policy *different ways of counting and defining claims Date of loss/accident date/occurrence date - date of the event that caused the loss or date when damage became apparent Report date - when claimant reports claims to insurer Report lag - the difference between accident date and report date IBNR/unreported - incurred but not reported (unreported) claims Settlement lag - time between report date and date of ultimate settlement (this and report lag distinguishes short vs. long-tailed lines) *once a claim is reported, is it considered open until settled and closed

How are exposures calculated for large commercial risks?

Composite rating: -Rate a large, complex policy with multiple coverages using a single exposure base, instead of rating each coverage with a separate exposure base -Calculate ORIGINAL policy premium with different exposure measures for each aspect of coverage (ex: sales revenue for general liability, property value for commercial property, etc.) -Instead of auditing each individual exposure measure, use a proxy to gauge overall exposure to loss (ex: property value up 20%, raise premium 20%) Loss-based composite rating: -Calculate premium based entirely on individual risk's historical loss experience (no manual rate) -Insured is eligible if historical reported losses + LAE over a defined perid exceed a specified aggregate dollar amount

Most common external data sources

Data calls/statistical plans (NCCI, ISO), other aggregated industry data (Fast Track Monitoring System), competitor rate filings, other third party data (economic & geo-demographic data like CPI and weather indices + census data and credit data) *geodemographic data = average characteristics for people of a particular area

What is a catastrophe? How do insurers count it in their claims data?

Defined by Insurance Services Office (ISO) as: 1) Results in at least $25 million direct insured losses to property 2) Affects significant number of policy holders and insurers Insurers usually count catastrophes as 1 occurrence because policies normally have coverage limits per occurrence (lots of litigation over whether 9/11 counted as 1 vs. 2 occurrences since there were 2 planes)

Rating manuals

Document containing the information necessary to appropriately classify each risk and calculate the premium associated with that risk Final output of ratemaking process is the information that allows creation or modification of rating manuals

Discuss case reserves

Estimate at the individual claim level of the total future payments required to settle that claim based on information available (done by claims adjusters, formula, or not at all) *not allowed to be TVM discounted (except for Workers Comp indemnity benefits which are discounted using NCCI table) *SOMETIMES includes separate costs for insurer (legal, etc.) *updated to reflect latest info.

What are some anomalies outside of losses & LAE and how do we process them in ratemaking?

Ex: purchasing new policy processing computer system for entire company *special accounting rules apply *approach 1: smooth out costs over time with a loading factor *approach 2: don't price for expense and pay for out of existing surplus (use when future policyholders probably shouldn't have to pay for the cost)

Explain why actuaries should be careful when comparing in-force premium for portfolios that write policies with different terms

Example: two insurers write the same volume of written premium (in terms of overall dollar amount), but one insurer writes semi-annual term policies and the other writes annual term policies On average, the in-force premium of the insurer writing semi-annual policies will be half that of the other carrier

Discuss the common fields in a policy database

Examples: 1 home for an annual policy period, 1 industry classification for a workers' comp policy, 1 coverage/1 car/1 driver/etc. for auto (varies) Policy identifier - Unique identifier (primary key) Risk identifier - Only used for products that insure multiple risks on a single policy Relevant dates - Original effective and termination dates (separately maintained for different risks/coverages) + midterm adjustments like change in deductible + transaction processed date Premium - Written premium; often separated into different coverage indicator fields Exposure - Written exposures; often recorded by coverage Characteristics - Rating and underwriting variables + all other information describing the risk **1 new record required for cancellation and 2 new records required for midterm adjustment of premium (one to close out old info and one to add new info)

What are the distribution channels for receiving a quote? How is a quote normally created?

Exclusive agent who sells for a single insurance company usually in exchange for a higher commission Direct, usually over the internet Independent agent who sells for multiple insurance companies and often becomes a product specialist (common in commercial lines ______________ Obtain risk characteristics from customer and third party sources (ex: DMV) then provide the quote usually with partial information (to avoid unnecessary fees and get it out quickly) --> can give refund later (ex: audit in WC)

Exposure & Premium: definition, types, & basic facts

Exposure: basic unit of risk underlying the insurance premium Written exposures/premium - total exposures/premium arising from policies issued/underwritten during a specified period of time (look at effective date & doesn't depend on valuation date) Earned exposures/premium - portion of written exposures/premium for which coverage has already been provided as of a certain point in time (what insurer gets back if cancellation) Unearned exposures/premium - portion of written exposures/premium for which coverage has not yet been provided as of a certain point in time In-force exposures/premium - number of insured units/full-term premium for policies that are exposed to loss at a given point in time (no consideration for the duration of the exposure i.e. 6 month policy DOES NOT count as 0.5 exposures) *can aggregate any of the above by CY or PY *written = earned + unearned (overall or for PY, not CY) *written = earned + change in unearned (always holds)

How does on-leveling written premium differ from earned?

For Policy Year: identical to earned For Calendar Year: DIFFERENT (COME BACK TO HOW B.3.5 Slide 2-3) example logic: 50 policies would have been written between 10/1/11 and 3/1/12 at premiums of $1000 each (arbitrary) so the 5% rate change represents a $50 increase in premium for each policy. On average these policies will be in-force for another 3 months on 10/1/12 so the written premium in CYXX for these policies from the law change will be 50 policies x $50 x (3/12)

Pure Premium/Loss Cost/Burning Cost

Frequency x Severity (numerator and denominator that cancel out need to be consistent) Losses / Number of Exposures *usually reported losses (or ultimate) and earned exposures

How do deductible shifts impact frequency & severity?

GENERALLY: OPPOSITE EFFECT Higher deductible --> lower frequency, higher severity, higher pure premium

ASOP 13 notes [repeated read through]

Guidance for P&C actuaries when trending *Forecast period is the future time period to which the historical data from experience period are projected *Economic influences like inflation *Social influences include costs of societal changes: changes in claims consciousness, court practices, legal precedents, etc. *Consider conflicts of interest arising from different purposes for trending, how estimate is presented (point estimate, range, etc.) When selecting data (could be historical insurance data or non-insurance data), consider: *credibility, time period, relationship to items being trended, the effects of biases or distortions like one-time events or changes in coverage or deductibles When selecting trend period, consider: *changes in mix of business between experience and forecast periods and timing of relationship between different data sources Document assumptions, procedures, methods, purposes, etc.

When might a weighted average be good for a problem? ]When might it be good to not use a calculated ultimate estimate from dev method?

If base (ex: earned exposures) is much larger in some years than others (esp. recent years) then consider taking an EXPOSURE WEIGHTED AVERAGE If ultimates are consistently very close to each other in all the prior years, might be best to judgmentally select prior years value if recent year looks off (or use recent year and state that perhaps there was an operational change causing it to be different)

How to deal with non-annual policy term for exposures? (same for premium)

If policies are six month for example, each policy represents HALF a written exposure and for the pictures, the policy line will be twice as steep

Written exposures facts and formulas (**SAME FOR PREMIUM**)

Individual policy & policy year aggregation at certain point in time (represented by start of line on bottom axis) **each policy only contributes WE to a single calendar/policy year, and can contribute to two CALENDAR years if a cancellation occurs and date of cancellation is different from original effective date (can't contribute to two POLICY years) --> negative written exposure/premium in the amount of the unearned exposure/premium as of the LATER OF (EFFECTIVE DATE OF CANCELLATION, CANCELLATION TRANSACTION PROCESSED DATE) [[the same logic applies for a midterm recalculation of premium]] Written exposures = earned exposures + unearned exposures (earned approaches written as time moves) Calendar year aggregation: CY Written Exposures = CY Earned Exposures + CY Unearned Exposures - Unearned Exposures as of beginning of CY

How do large events & anomalies, One-time changes, and development impact trend?

Large events & anomalies: *distort the true trend rate (ex: COVID) *remove catastrophe data in both getting the trend & applying --> load for catastrophes separately One-time changes: *on-level i.e. restate at current rate levels before trending Development: *NO OVERLAP BETWEEN DEVELOPMENT AND TREND *development brings immature data from each historical period to its ultimate levels, while trending reflects the difference in ultimate levels from one period to the next i.e. makes sure ultimate losses are at cost levels corresponding to the future policy period

LAE & subcomponent definitions

Loss Adjustment Expense - claims-related expenses, incurred in the process of investigating, adjusting, and settling claims LAE = ALAE + ULAE = DCC + A&O (introduced in 1998 in an attempt to improve financial reporting consistency between insurance companies -- companies with in-house attorneys sometimes coded certain expenses as ULAE, creating differences) Allocated: directly attributable to specific claim (ex: legal fees to defend against specific claim, costs incurred by claims adjuster assigned to one claims) --> often included with losses for ratemaking purposes Unallocated: can't be directly attributed to specific claim (ex: claims department salaries) Defense Cost and Containment expenses: cost incurred in defending claims (expert witness fees and other legal fees) Adjusting and Other expenses: all other expenses *ALAE & DCC vary by dollar amount of each claim, while ULAE + A&O vary by number of claims reported

Combined Ratio

Loss Ratio + LAE/Earned Premium + Underwriting Expenses/Written Premium = Loss Ratio (1 + LAE Ratio) + UW Expense Ratio (according to TIA) Loss Ratio + OER (if company uses earned premium in UW expense ratio) *portion of each premium dollar used to pay for loss adjustment and underwriting expenses --> CR = 98% means that 2 cents of every premium dollar goes to UW profit

Loss definitions

Loss: AMOUNT paid or payable to claimants (insured or third-party) under insurance contract Claims: demand for compensation by claimant Paid loss: loss that has been paid to claimants Case reserve: established when claim is reported and is estimated amount of money required to ultimately settle claim *ultimate loss - amount of money required to close and settle all claims for a defined group of polices *IBNR - reflects unreported claims (includes claims in transit) *IBNER - development on known claims & reflects unreported information --> difference between amount estimated to ultimately settle reported claims and aggregate reported losses when losses are evaluated (ex: claim is reopened) *IBNR (broad definition): includes IBNR and IBNER

Severity

Losses / Number of Claims *Paid severity uses paid losses on closed claims and closed claims *Reported severity uses reported losses and reported claims *ALAE may or may not be included and losses may or may not be developed to ultimate

How can competitor or industry/bureau rates be made more applicable to book being priced?

Make some adjustments (factor depends on whether figure is in numerator or denominator of calculations): Fixed Expenses: Adjust competitor expense fee by a factor of (Your Fixed Expense Per Exposure/Competitor Fixed Expense Per Exposure) Variable Expenses: Adjust competitor base rate and expense fee by a factor of (Competitor Variable PLR/Your Variable PLR) Loss Costs: Adjust competitor base rate by factor of (Your Expected Loss Cost / Competitor Expected Loss Cost) Classes: judgmentally adjust class relativities to target different market segments

What adjustments need to be made to historical aggregated loss data in order to estimate for a future period?

Non-recurring extreme events: shock losses, catastrophes, and benefit changes, Develop to future settlement levels and trend to future cost levels Add in LAE

Close Ratio

Number of Accepted Quotes / Number of Quotes *AKA hit ratio, quote-to-close ratio, conversion rate

Frequency

Number of Claims / Number of Exposures *usually reported claims and earned exposures

Retention Ratio

Number of policies renewed / number of potential renewal policies *gauge competitiveness of rates and used in projecting future premium volume

What are some of the most important considerations in addressing an imbalanced fundamental insurance equation?

OVERALL: any data is relevant if it helps predict future costs *steady state environment: equation will eventually return to being balanced (ex: underwriter intentionally underestimates exposures, rates will gradually rise to bring premium back to a level that balances fundamental insurance equation) *consider uncertainty in expected costs: more risk/uncertainty --> demand higher target UW profit *consider adjustments, reinsurance, policy provisions, investment income (ex: get more investment income form long-tailed lines), credibility (ex: detailed segmentation increases homogeneity but reduces credibility)

Discuss exposure trends

Only applied to lines w/ inflation sensitive exposure bases (ex: payroll for workers comp) Represent change in AVERAGE exposures per policy due to inflation Inflation usually causes total premium & losses to increase ceteris paribus

Expense trends

Only care about FIXED (not variable) Represent change in average fixed expenses over time due to inflation *if incurred at start of policy: trend using policy effective (written) dates (average written date for WP) *if incurred throughout policy (general expenses): trend using average earned dates (average earned date for EP)

Discuss the major P&C lines & their typical exposure bases

P&C Insurance divided into Personal & Commercial Personal: *personal auto - earned car years *homeowners - earned house years *personal articles floater - value of item *personal umbrella Commercial: *commercial auto *commercial property - amount of insurance coverage (for business property) *commercial general liability - sales revenue, payroll, square footage, number of units *professional liability - number of professionals (for lawyers, accountants, etc.) or number of physician years (for doctors) *workers comp - payroll [read through Section A.1 PDF & be able to define]

Discuss the common fields in the claims database

Policy identifier - Matches with policy database Risk identifier - Matches with policy database Claim identifier - Unique ID (could be multiple claim transactions) Claimant identifier - Unique ID for each claimant on a specific claim (could be multiple) Relevant loss dates - Accident date, report date, & dates of loss payments, reserve changes, claim status changes, etc. Claim status - Open/closed or re-opened/re-closed Claim count - # of claims by coverage associated with the loss occurrence Paid loss - Often tracked separately by type of loss or coverage & separated by CAT Event identifier - Identifies extraordinary event like a CAT Case reserve - Often tracked separately ...^ & typically updated when a paid loss happens ALAE- Often tracked separately ...^ Salvage & Subrogation - Reconditioning damaged property (salvage) & recovery damages from another liable third party (subrogation) Claim characteristics - Associated with the claim (type of injury, physician info, etc.) *only characteristics known for every policyholder at the time of quotation can be used in the rating algorithm*

Discuss Excess Loss Loading and considerations when calculating it

Potentially part of the adjustment for shock losses --> Excess Loss Factor = 1 + Losses in Excess of Cap Level/Losses Below Cap Level (usually an average ratio over 20-30 years i.e. take ratio for each year THEN average) --> multiply non-excess losses by ELF for the years used in pricing (3-5 years usually) ***ELF FACTOR = 1 + ELF RATIO *to account for change in severity levels over time (inflation) 1) Cap based on future policy period cost levels & trend historical losses to this cost level 2) Have cap level change by year to reflect inflation --> alternative: fit statistical distributions and run a simulation

Average Premium

Premium / Number of Exposures *premium and exposures should be on the same basis

Fundamental Insurance Equation & UW definitions

Premium = Losses + LAE + UW Expenses + UW Profit *goal of ratemaking is to determine adequate future rates that will produce premium to balance this; requires estimating each component for a future period Underwriting Expenses: operational/administrative expenses incurred in the acquisition and servicing of policies (commissions and brokerage, other acquisition, taxes, licenses, fees, etc., and general) Underwriting profit/operating income: compensation for risk (other form of profit for insurance companies is investment income)

What is the typical way premium and losses are aggregated?

Premium by Calendar Year (don't want to wait 24 months for policy year to complete) Losses by Accident Year

What premium data and loss data do we use for getting TREND RATE for trending?

Premium data: *if forecasting loss ratios (ex: for indications), use EARNED PREMIUM *average OLWP can be a leading indicator for EP and may be used to forecast *can use CY (including or excluding audits) or PY ultimate avg. premium Loss data: *usually freq/sev, sometimes pure premium (usually do (1+freq_selected)(1+sev_selected)-1 for pp) *short-tailed lines: use CY paid or reported data (assumes book not significantly growing or shrinking and ignores mismatch) (paid not subject to reserving, but reported captures more info) *long-tailed lines: use AY ultimate data (ultimates must be estimated which compounds uncertainty)

How do ratemaking and reserving differ?

Pricing actuaries: estimate cost + target profit of policies before they're sold *balance at individual/group level Reserving actuaries: estimate cost for policies after they're sold --> solve estimated profit = price - total cost *balance at LOB AY level

Proportional vs. Non-proportional Reinsurance

Proportional covers: same proportion of premium and losses are transferred/ceded to reinsurer (may not need to explicitly include in pricing consideration) Non-proportional covers: insurer cedes a portion of premium to reinsurer, who covers some predefined portion of the loss (the reinsurance recoverables) 1)Catastrophe excess-of-loss reinsurance (e.g., the reinsurer will cover 50% of the losses that exceed $15,000,000 up to $30,000,000 on their entire property book of business in the event of a catastrophe) 2)Per risk excess-of-loss reinsurance (e.g., the reinsurer will cover the portion of any large single event that is between $1,000,000 and $5,000,000 for specified risks).

When might it be necessary to restate historical experience?

Rate changes, operational changes, law changes, inflationary pressures, changes in the mix of business written

Trending for layers of loss

Recall example: "2% annual ground-up severity trend" --> predict FULL loss amounts will increase by 2% each year Trend rate for LIMITED LOSSES is generally less in magnitude (less positive/negative) than the GROUND-UP trend rate Trend rate for excess losses is generally greater than the total or basic limit severity trend 1) For losses above basic limit, entire trend is in XS (excess layer) 2) Losses just below basic limit get pushed into XS layer Formula: not taking average of percent changes Ex: excess loss severity trend = (EXCESS LOSSES AFTER APPLYING TOTAL LIMITS SEVERITY TREND /EXCESS LOSSES WITHOUT APPLYING TREND) - 1 & basic severity trend = basic trended loss/basic loss - 1 (where basic loss is loss below basic limits)

What are the types of recoveries that insurer can receive?

Salvage: proceeds from sale of damaged property Subrogation: recovering costs from a liable third party Deductible recoveries Reinsurance: percentage or in layers

Report Year Aggregation

Same as (calendar) accident year except losses are aggregate according to when claim was reported (not when claim occurred) *useful for measuring lag and estimating IBNER (NOT IBNR) *results in no IBNR and a shortfall in case reserves --> NOT GOOD WHEN LOOKING AT OCCURRENCE POLICES THAT WILL HAVE IBNR (EX: AUTO LIABILITY) *good for isolating change in claims practices (ex: case reserve adequacy) *typical for commercial lines products using claims-made policies (ex: medical malpractice)

ASOP 43 notes [repeated read through]

Statement of Principles Regarding P&C Loss and Loss Adjustment Expense Reserves/P&C Unpaid Claim Estimates (might be worded like this in exam) Guidance to actuaries when estimating unpaid claim estimates (includes loss + LAE): · Consider intended use of estimate (pricing, reserving, business decision, etc.) and potential conflicts; consider purpose of estimate in deciding how to present it · Communicate and document constraints (limited data or resources) that could materially impact estimates · Identify intended measure (high estimate, mean estimate, etc.), discounting used (TVM), gross or net of recoverables (S&S), types of LAE used, claims to be covered (type of loss, LOB, year, state), methods used in prior analysis, impact of external conditions, etc. · Actuary may disregard immaterial items depending on scope · Actuary should be familiar with the product/LOB they are working with and its nuances · Actuary should check estimate for reasonability · Actuary should document limitations risks and uncertainties ALWAYS CONSIDER HOMOGENEITY AND CREDIBILITY OF DATA

List the types of adjustments to historical data that are needed in ratemaking ***IMPORTANT

Statement of Principles Regarding P&C Ratemaking Large events & anomalies One-time changes (ex: legislation) Continuous changes Development Loading for UW expenses & ULAE Setting UW profit target Reinsurance costs Credibility

Why is accurate estimation of unpaid claims (reserving) important?

Suppose estimates of unpaid claims were inaccurately high, for example... 1) Internal management might unnecessarily raise rates, tighten underwriting guidelines, exit a LOB or territory, or purchase additional reinsurance 2) Investors might think that the insurer is less profitable i.e., a worse investment with worse returns 3) Regulators monitoring insurer solvency might restrict the insurer's ability to write new business 4) Laws might be violated, as proper estimating is required by local laws: most P&C insurers must obtain a Statement of Actuarial Opinion signed by an qualified actuary, overseen by an Appointed Actuary

Large Deductible Policies: overview, pricing considerations, pricing method

Typically for policies purchased by large employers (ex: $100k+ per occurrence) Unique considerations: *Claims handling: is insurer or insured responsible for adjusting claims below the deductible? If former, usually insured will higher TPA and needs to be careful because insured will have less incentive to keep losses below the deductible *Application of deductible: could apply to losses only or losses + ALAE *Deductible processing: extra costs and credit risk if insurer has to seek reimbursement from insured for deductible *Risk margin: usually higher because losses above the deductible are volatile and difficult to estimate ***Pricing method (important): *use Pure Premium formula *for numerator, use expected losses in excess of the deductible and for ALAE, depends on whether it's subject to deductible (use excess of deductible) or not (use ground-up amount)

Operating Expense Ratio

UW Expense Ratio + (LAE / Earned Premium) = Combined Ratio - Loss Ratio *portion of each premium dollar used to pay expenses and LAE

What is total profit for an insurer?

UW Profit/Operating Income + Net Investment Income

Underwriting Expense Ratio

Underwriting Expenses / Premium *subdivide the expenses into categories *expenses incurred at onset of policy: use written premium *expenses incurred throughout the policy: use earned premium

What is experience rating?

Use an insured's claim history on prior policy terms in determining the current policy premium Produce an experience modification factor (experience "mod") by credibility weighting an insured's expected loss experience with their actual loss experience Experience mod (sometimes subject to max or min values) is applied to manual premium

Discuss underwriting guidelines

Used by insurer to decide whether or not to offer coverage (usually kept secret) *accept, decline, or refer risks *company or tier placement (put customers into different writing companies) *schedule rating

When is two-step trend used?

Used when a company expects premium trend to change over time i.e. trend during historical period has been significantly different from what is expected to occur in the future (ex: homeowners insurance when insurer observes significant decreases in AOI during experience period that are not expected to continue into the future) *same idea for loss trend: another example would be if there was a legislative change during the historical period

Discuss accounting data and the different types

Usually highly summarized and actuaries generally want more granular data Types: *summarized policy and claim data by LOB and various years/periods *reinsurance data applied to groups of policies *ULAE allocated to LOB (tricky to do sometimes, ex: CEO salary allocated to LOB by earned or written premium) *investment data - income and expenses *underwriting expenses Financial Reporting: Insurance Expense Exhibit - all expenses allocated by LOB Schedule P - all ULAE allocated by accident year

Why do we adjust for one time changes and what effects should we consider?

Want to restate historical data to be reflective of future rate levels, coverage levels, and expense levels Direct effects - impacts to prem/loss/expenses from changes Indirect effects - impacts to prem/loss/expenses from changes in human behavior that are due to the one-time change (OFTEN INCORPORATED INTO TREND ADJUSTMENTS) *ex: increase premium --> customers leave and lower close and retention ratios *ex: WC indemnity benefits increase --> more workers file claims bc higher incentive to stay home

What are distributional changes and give a few examples:

When average premium level changes over time due to characteristics of the policies written (not things like inflation) --> resulting change in average premium level is called premium trend Examples: 1) Rating variable changes (ex: homeowners premium depends on AOI which is indexed by inflation) 2) Company decides to move all existing insureds to higher deductible 3) One company purchases the entire portfolio of another company

When is it difficult to apply one-step trending?

When changes in average premium vary significantly year-by-year and/or When the historical changes in average premium are significantly different than the changes that are expected in the future (ex: one time shift to higher deductible) SOLUTION: two-step trending

When might one want an exposure trend and how does it work/what does it affect?

When exposure measure is sensitive to time-related influences such as inflation (payroll for WC and sales/receipts revenue for GL example) which can impact fundamental insurance equation and the overall rate indication Use historical exposures to project future exposure (AND THUS ALSO PREMIUM) to future inflationary levels Estimate these trends via internal insurance company data (ex: workers comp payroll data) or via industry & government indices (ex: average wage index)

When is 15th of the month rule / 24ths method not a good approximation?

When policies are NOT written uniformly during each time period (can be improved by using shorter time periods if possible)

Why use 2 step trend instead of 1 step? What are the steps of Two-Step Trending?

When the historical trend rate differs from the expected future trend rate Apply the trend in 2 steps: 1) Current (historical) trend to latest data point 2) Future (prospective) trend from latest data point to future period ***for two-step loss trends, typically just use average accident dates for all three points*** *same process as One-Step trending if using method #1 for current (see notecards)--> just insert breaking point as midpoint of latest data point (ex: 11/15/21 for 2011 Q4 or accident quarter 11Q4 for losses)

WP vs EP for calculating premium trend?

Written premium is a leading indicator of trends that will eventually emerge in earned premium --> usually apply trends observed in WP to historical EP Actuaries normally use quarterly average WP to make the statistic as responsive as possible

ISO P&C unit definition of catastrophe

events that cause $25 million or more in direct insured losses to property and that affect a significant number of policyholders and insurers

What is reserving? Give all relevant loss equations

the process of evaluating, reviewing, and estimating unpaid claims ______________________________ *let "claims" denote dollars of loss in this context (not claim counts) *note that the following equations can also apply to estimating losses *"reported" = "case-incurred" ULTIMATE CLAIMS = PAID CLAIMS + IBNR (BROAD DEFINITION: IBNER + IBNR) + CASE RESERVES Reported/Case-Incurred Claims = Paid Claims + Case Reserve --> ultimate = reported + IBNR (broad) Unpaid Claims/Total Claims Reserve = IBNR (broad def.) + Case Reserve --> ultimate = paid + unpaid ***ULTIMATE PAID = ULTIMATE REPORTED (case reserve will eventually fall to zero) ***CUMULATIVE REPORTED CLAIMS = CUMULATIVE PAID LOSS + ENDING CASE RESERVE