Triangulation methods incomplete Flashcards
What does triangulation mean?
- To review the development of claims/premiums/other quantities over time
- This contrasts to aggregating all the claims/premiums/other quantities for a given period of time
Earned reserves defn.
Reserves held to cover the expected cost of all claims that have occurred whether they have been reported to the insurer or not
- Earned premiums is a starting point for finding earned reserves
Unearned premium reserves defn.
- Proportion of paid premiums kept aside to pay the cost of claims likely to emerge on unearned exposures
- If these reserves are seen as insufficient, additional unexpired risk reserves (AURR) may be added to them
Reserving exercise done on an accident year basis produces what reserves?
Earned Reserves
- Because it considers accident years and paid claims in years after
- Resulting reserves are for claims that need to be paid plus those that have not been reported yet (IBNER + Pure IBNR)
Reserving exercise done on an underwriting year basis produces what reserves?
Earned reserves + Unearned reserves
Explanation:
Reserves meant to cover the cost of claims that have already occurred (both reported/not reported) + those likely to arise on the unearned exposure for unexpired policies @ the end of the year which were written in the U/W year of interest
Best estimate reserves
- Generally understood as the mean of the dbn of ALL POSSIBLE OUTCOMES of unpaid liabilities
- Possible alternatives may be:
o The median
o The mode
o Estimate w particular prob. of exceeding the outcomes
ALL POSSIBLE OUTCOMES in context of unpaid liabilities meaning:
- Includes all reasonably foreseeable events based on exposures of the insurer
- Under solvency II, it also covers new claim types that the insurer has never experienced before called “events not in the data” or ENIDs
Describe the dbn. shape of unpaid liabilities? Why does it look like that?
- Distribution is typically positively skewed
- This is due to high severity low frequency claims experienced in GI
- This results in the MEAN being larger than the median
- This makes the MEAN a more prudent estimate of best reserves
What does the timing of reserving exercises depend on?
- The needs of the business
o Estimate reserves quarterly & review assumptions annually unless experience indicates need to review assumptions for frequently (COVID) - Capacity of the reserving team
- Usually aligned to when the accounts are published (annual or bi-annual reviews)
Uses of reserving exercises within the company:
MAIN:
- To set provisions to be booked on the balance sheet of the insurance company
OTHER:
- Assist business in setting BUSINESS PLAN
- Identify key areas of UNCERTAINTY in portfolio of business (volatile reserves)
- Valuing the business for purchase or sale (M&A)
- PARAMETERISE the reserve risk dbn used for capital modelling
- Identify key areas of business that are underperforming or underpriced (SLACKING)
- Identify trend in CLAIMS HANDLING PRACTICES (slowing down or speeding up depending on run-off time)
- Starting point to determine CAPITAL REQUIREMENTS for insurance risk
BUMPS CC
What is Actual/Expected (A/E) analysis about?
- Comparing the actual experience against the expected for quantities such as:
o Paid/incurred claims
o Number of claims
o Average new claim size
o Development of existing claims
- Quantification of the differences in A&E ie. what proportion of the difference observed in A&E was due to:
o Experience being different to expected
o Changes to actuarial assumptions/methodology
o Changes to historical data (error/reopened claims)
o Foreign exchange movements
What may cause differences in actual vs expected experience?
o Experience being different to expected
o Changes to actuarial assumptions/methodology
o Changes to historical data (error/reopened claims)
o Foreign exchange movements
What are the main claim cohorts to consider when reserving?
- Accident year
- Underwriting year
- Reporting year
** In reality we may actually use accident quarter or month
What consideration is most important to make when using a claim cohort of a particular type for analysis of claims development?
- Consistent bases should be used for both claims and the exposure to risk
- Exposure and claims being analysed must be from the same analysis period and claims must include both settled/unsettled
What are two possible definitions of incurred claims?
- Paid claims plus estimates for outstanding reported claims e.g. case estimates
- Paid claims plus all estimates for outstanding claims (including IBNR loading)
Development period defn.
- This is the period or frequency at which each claim cohort is tracked over time e.g. yearly, quarterly, monthly
- Usually quarterly or monthly
o Management report: usually quarterly - Does NOT need to be the same as cohort period
Accident year claim cohorts description:
- Claims are grouped according to the year in which the claim event occurred
- IBNR, reopened claims & recoveries will be allocated to the original accident year cohort in which they occurred
- Projection of future claims development (reported or paid) should allow for IBNR claims, reopenings & recoveries belonging to the cohort
- Exposure base used is earned premiums
Advantage & disadvantage of accident year cohort:
Adv.
- All claims stem from the same exposure cohort => claims will have usually been subject to the same risk environment BUT claims arise from policies that were underwritten under different rating & policy terms
- Accident year corresponds w accounting year => able to compare the actual experience for the year w what was booked in the accounts
Dis.
- Full number & amount claims in the cohort is unknown until the last claim is reported & settled
- Date of loss is not always known e.g. subsidence or asbestos for employers liability -> very gradual
Reporting year cohort description:
- Claims grouped according to the year they were reported to insurer/reinsurer regardless of the original period of claim event or inception
- Reopened claims should be allocated with care depending on the situation
- No natural exposure base, usually average premium or current year’s premiums are used as proxies
Advantages & disadvantages of reporting year:
Adv.
- No further claims added to the cohort after the end of the reporting period -> able to assess the delay before reliable estimates can be observed for notified claims
Dis.
- No allowance for IBNR -> separate allowance for IBNR needed
- Claims come from several different exposure periods eg. volume of business, cover applying & claim settlement patterns
- Does not pick up on changes in exposure or risk profile
Underwriting year claim cohort description:
- Claims grouped according to the year in which the policy covering the claim was incepted regardless of when the claims were incurred
- IBNR claims, recoveries & reopenings are automatically included provided they are all allocated to the correct U/W year
Advantages & disadvantages of underwriting year cohort:
Adv.
- Can track total outcome of policies written in each year
- Claims subject to the same set of premium rates test the adequacy of premiums
- Terms, rates & conditions are normally more stable
Dis.
- It will take more than one year before the claims in the U/W year cohort are run-off
Why may some classes of business use shorter or longer development periods for cohorts? e.g. quarter, month
- Longer development periods usually more stable
- > Development may be volatile, we use long dev. periods and use linear interpolation to obtain required granularity level
- Shorter development periods (monthly) allow us to understand trends at different stages of development
-> Observe seasonal trends for products
o Offered on a seasonal basis eg. travel insurance for summer vacations
o That incur claims more heavily in certain seasons
-> Shorter development periods may make it difficult to identify bigger picture trends
o May need to apply judgment to smooth results for analysis
How to prioritise reserving work and considerations to make for various business classes when working on reserving:
- Focus on the more material groups/classes (in terms of size & uncertainty)
o Any significant issues can be identified earlier rather than later
CONSIDERATIONS:
- Materiality also affects the selection of methodology
o Obtain more accurate projection for material tasks - Classes w small premiums/past claims may actually have large underlying exposures
o These classes need more analysis than suspected - Certain classes are tough to project using standard actuarial approaches e.g. Excess treaty for liability classes w high attachment levels have lack of data, longer reporting delay/settlement period
o Such classes result in no loss or large loss
o Use freq-severity analysis or exposure based method
** freq-severity analysis: Set a dbn which models severity of claims which varies w freq. and decide at which point the particular class is represented on the dbn curve
** exposure based method: Use an exposure benchmark (internal or external) and adjust the benchmark for our own portfolio
Considerations for data heterogeneity:
- Need high quality data (always important)
- Different needs in terms of quantity & type of data:
o If data is needed on big picture basis -> publicly available statutory returns & published accounts may suffice
o Reserving needs more detailed data - Always balance homogeneity & credibility
Challenges presented by data heterogeneity:
- Management might want results on a product type, scheme, dbn channel or even individual broker (finer level of detail) =>
o This goes against grouping data for required homogeneity
o Time consuming => insufficient analysis of data & overly-mechanical approach to meet deadlines - Hence reserving is done at a more desired aggregated level
o Then allocation approach used to align IBNR against required level of granularity - If the data is impossible to subdivide into homogeneous groups:
o Ensure the mix of diff. subdivisions within the data is reasonably stable through time for stable aggregate development pattern
How should large losses and non-standard risks be analysed?
- They should be analysed separately from attritional claims
o They may develop faster than attritional claims e.g. due to catastrophe losses get extra attention from loss adjustors & p/hs
o They may develop slower than attritional claims due to court disputes - In early stages of development, credibility-based approaches used (there is very little info available)
o A benchmark may be used to estimate ultimate claims - As more info emerges, usual exposure based & claims development methods are used to find ultimate claims
o Large claims are projected from the date of actual loss and not the start of the origin period when using claim development method
o Exposure based method has two approaches, bottom up and top down - Need to decide whether to cap losses or not
o May cap if we decide claims above a certain amount are 1 in 200 year events (don’t need to reserve for it under Solvency II)
o Need to adjust cap amounts for inflation
Large losses exposure based methods: Bottom up
Bottom up
- Project the losses at policy level first & sum up losses to get overall reserves
- U/W staff/Claims staff estimate the likelihood of each policy being exposed to the loss event considering:
o Dates
o Perils covered
o Location
- If likely, claims expert estimates claim size of the
policy considering:
o Size of loss in relation to excesses & limits - Binding authority business (sold & signed by brokers) usually provides limited info on underlying exposures hence leading to underestimation
Large losses exposure based methods: Top down
- Estimate the total market loss first
o For coinsurance: total mkt loss = sum of provisions of each coinsurer (w diff views/risk appetites) - Then allocate the loss to each class/policy group based on terms of policy, excesses & limits
