Objective 3 - Reserving Flashcards
Additional considerations in establishing claim reserves
(in addition to ASOP #5 considerations)
- Incurral dating method
- Reserve basis - statutory, GAAP, and tax bases differ in their use of margin, interest rates, etc.
- Interest - reserves for claims with long payouts may be discounted to reflect interest
- Controls and reconciliation - the data used should be tested for accuracy
- Insurance characteristics - reserves vary depending on the type of risk covered
- Reserve cells - set up separate cells for each homogenous category of business
- Managed care features - such as discounts and provider risk sharing arrangements
- Trends
- Seasonality
- Claim administrative expenses - set up a reserve equal to a percentage of the claim reserve
- Morbidity assumptions - for long-term claims, morbidity is reflected in continuance tables
- Use of the case reserves method - very labor intensive, so only recommended for small blocks
Advantages and disadvantages of stochastic approaches to reserving
Advantages
- Provides explicit guidance for establishing provision for adverse deviation in the reserves
- Provides guidance on potential variability in reported earnings and reserve levels
- Allows for quantification of variability in items such as seasonality and claim trend
- Allows for improved evaluation of reserve estimates (by knowing the variability of the estimate)
Disadvantages
- Some audiences that are unfamiliar with this approach may have a false sense of confidence in the approach because of its sophistication
- May be too complex to be used by all individuals who must perform related functions (such as forecasting and pricing)
- Not every process can be modeled rigorously
Stochastic modeling techniques for reserving
- Fitting a parametric distribution to the data - this technique works best when the process being modeled is stationary over time
- Ordinary least squares regression - this allows for investigation of the effects of specific explanatory variables, such as trend or seasonality
- Generalized linear models - these models improve upon ordinary regression models because they allow for cases where the dependent variable being modeled is either bounded (e.g., must be greater than zero) or not normally distributed
- Stochastic time series models - these are useful for handling situations where values are correlated across time (e.g., seasonal or cyclical patterns)
- Monte Carlo simulation - this approach is of significant practical value when combining results from any of the other techniques
Considerations when developing a stochastic approach to reserve estimation
- Availability of data - historical data is needed to validate the model and assumptions
- Appropriateness of data - consider whether the process reflected in the historical data are representative of the process being modeled going forward
- Access to statistical software - lack of access to or understanding of modeling software will limit the available choices for modeling techniques
- Appropriateness of the model - this can be validated through goodness-of-fit testing, residual analysis, and hold-out sample evaluation
5, Covariances of modeled estimates - when reserve estimates are calculated through component estimates, the covariance between these components must be estimated
Features (or aspects) of LTD and LTC contracts to consider when setting reserves
- Periodic benefits - benefits typically equal some specified monthly or daily amount
- Long-term benefit periods - these plans have maximum benefit periods that are much longer than benefit periods for other health benefits
- Elimination periods - LTD and LTC plans have a variety of elimination periods (often 90 days or more)
- Optional benefits - these may affect the timing or amount of monthly payments (e.g., partial disability benefits and cost of living adjustments)
- Integration of benefits - these plans often coordinate benefits with Social Security and Medicare
- Limitations and exclusions - some claims are excluded (such as intentionally self-inflicted injuries) or subject to limited periods (such as mental and nervous claims)
Types of long-term claims and reserve methods
- Open claims - claims currently being paid (uses tabular reserves)
a) Reserve = Vn = SUM(t=n to BP) of (Benefit,t * Continuance,t * InterestDiscount,t)
i) Benefit = monthly benefit (may vary over time due to product provisions)
ii) Continuance = probability of a claim continuing to receive payments in the future
iii) Interest discount = factor to reflect time value of money
iv) Summation runs from current time period (n) to end of benefit period (BP) - Pending claims - claims that have been reported but payments have not yet begun
a) Reserve for claims that are still in the elimination period = pending factor * tabular reserve
b) Reserve for claims that have completed the elimination period = pending factor * (tabular reserve + accumulated value for past payments not yet made) - IBNR claims - claims that have been incurred but have not yet been reported to the company (see separate list for reserve methods)
Methods for calculating IBNR reserves for long-term claims
- Percentage of premium method (a special case of the factor method)
a) For a historical year in which all claims have been reported, list all claims incurred prior to year-end that were reported after year-end. These are the IBNR claims
b) Calculate a tabular reserve for each of these claims as of year-end of the historical year
c) Sum these tabular reserves to get the IBNR reserve for the historical year
d) Divide this reserve by earned premium for the historical year to get an IBNR reserve factor
e) Multiply the IBNR reserve factor by earned premium from the current year to produce the IBNR reserve for the current year - Lag method (development method)
- Loss ratio method (see separate list)
- Combination methods - for example, use the lag method for earlier incurral months and the loss ratio method for recent incurral months where the completion factors are low
Common data integrity errors related to claim reserving
- Missing data
- Misstated age or gender
- Inaccurate elimination periods or benefit periods
- Incomplete or inaccurate information on benefit integration
- Inaccurate or inconsistent determination of the incurral date
- Inaccurate information on cause of diability
- Incorrect coding of claim status (open, closed, or pending)
Methods for evaluating claim reserve adequacy
- Runoff studies (commonly done by incurral year) - premium reserve balances are compared to subsequent claim payments and reserve balances, with adjustments for interest
- Actual to expected claim termination rate studies (commonly done by claim duration) - compares the actual claim terminations to the expected claim terminations based on the table used for reserving
- Experience studies - typically involves a gross premium valuation. The reserve is adequate if PV of future gross premiums + reserve > PV of future claim cost
(note that the 2nd point is a specific type of the experience studies mentioned in the 3rd point)
Duties of the actuary regarding the quality of data
- Seek out and use appropriate data and communicate any imperfections
- Review data for reasonableness and consistency (not necessarily an audit)
- Disclose reliance upon other for a review, reconciliation, or audit of the data
- Disclose situations where it is impossible or impractical to perform a sufficient review of the data
- Consider whether the use of inappropriate data might create a material bias in the work product
- Maintain adequate documentation to support the use of specific data
- Address reconciliation of paid claims to check registers or general ledgers
Types or reserve reporting
- Regulatory reporting - concerned with solvency and policyholder protection, so conservative
- GAAP reporting - emphasis on realistic earnings. Assumptions include provision for adverse deviation
- Experience reporting for employers and providers - typically less sophisticated except for financial settlement and pricing reviews. For settlements, allow a 3 month run-out period to minimize the size of the estimated reserve.
- Valuations for acquisitions - reserves are material to profitability, so they are often a focal point of negotiations. There is often a final settlement after several months to revisit the purchase prices and assess the impact of claim reserves
Types of claim liabilities and reserves
- Due and unpaid claims - liabilities that have been reported, adjudicated, and processed, but not paid. Is usually small. Itemize or base on historical averages
- In course of settlement - liabilities for claims reported and received but not year adjudicated and paid. System may record receipt and run report. Otherwise, use simple method such as average claim times number of claims.
- IBNR - liabilities for claims that are anticipated but have not been reported. Project by using existing payment data to develop average expected claims or claim payment patterns
- Loss adjustment expenses - liabilities for the administrative costs of adjudicating unpaid claims. Usually a percentage of unpaid claims liability
- Present value of amounts not yet due (“unaccrued”) - an estimate of future amoutns due to known open claims (commonly for disability or LTC claims). Usually reserved on a seriatim basis using a tabular approach
- Resisted claims - includes claims for which a known litigation situation exists. Usually reserved seriatim assuming full benefits and possibly amounts for damages.
- Outstanding accounting feed (may overlap with due and unpaid liability) - amounts which have been acknowledged as payments, but for which no check has been cut. Reserves are often based on accounts payable and billing notices.
- Deferred maternity or other extended benefits - the loss is triggered before the valuation date, but benefits are deferred by contractual provisions
- Other special reserves - such as for waiver of premium due to disability
Methods of estimation for claim reserves
- Case reserves - direct enumeration on a claim-by-claim basis. Typically used only when there are very few claims. Can’t use for IBNR.
a) Examiner’s method - estimate ultimate payment and deduct what’s already been paid
b) Average size claim method - the number of reported claims times an average claim amount minus the amount already paid - Projection method - develop a historical incurred claim rate as a function of some measure of exposure. Then apply this rate to projected exposure to get current incurred claims, and subtract claims already paid. The most common approach is: projected PMPM claim costs * member months - claims already paid.
- Loss ratio method (aka claim cost method) - loss ratio * earned premium - claims already paid. This method and the projected method can be used when the volume of data is small or to validate other methods.
- Tabular method - apply a factor to open claims to calculate reserve. Typically used for LTC and disability. Can’t use for IBNR.
- Development method (aka lag, completion, or triangulation method) - projects historical claim lag pattern into the future to estimate the reserve based on experience data
- Factor method (aka formula method) - historical studies are done of reserves paid after the valuation date for claims incurred before that date. These past reserves are stated as a percentage of some unit of exposure in the past time period (such as annual premium in force) to develop a factor. Current reserves equal this factor multiplied by the current amount of exposure.
- Stochastic approaches - methods where a probabilistic statement can be made about the level and adequacy of the reserve amount. Any of the methods discussed previously can be given a stochastic treatment.
Types of coverages for which the development method works well
- Ability to record incurral and payment dates for each claim
- Fairly consistent lag patterns
- Short incurral periods relative to the ultimate run out (monthly is preferred for medical)
- A sufficient volume of business in each cell, in order to obtain reasonably stable results
- Availability of either earned premium or exposure data (for volume adjustments and smoothing)
Steps of the development method
- Summarize the data by incurral month vs. paid month to get a claims triangle
- Sum the cells of the first claims triangle to get cumulative paid claims by incurral month
- Calculate age-to-age development factors as the ratios of month to month cumulative claims
- Smooth the month-to-month variations in age-to-age development factors. Various methods are used to do this (see separate list)
- Calculate age-to-ultimate development factors (called completion factors) from the smoothed age-to-age factors
- Divide each incurral month’s cumulative paid claims by its completion factor to get fully incurred claims
- Subtract cumulative paid claims from the fully incurred claims to get the unpaid claims liability
Smoothing methods to apply to development factors
- Simple averaging - average development factors for each lag month (3 month average is more current; 12 month average is smoother but may bury trends in the payment patterns)
- Removing bumps - throw out the high and low factors and average the rest. May also remove large “shock” claims from the claim triangle and analyze them separately.
- Weighted averaging - give more credibility to most recent results. Approaches include sum of digits, squared sum of digits, and constant declining percentage.
- Other types of means - harmonic (use the reciprocal of the “mean of reciprocals”) or geometric (the nth root of the “product of n observations”)
- Dollar-weighted methods (prior methods have been ratio weighted) - average cumulative payments for consecutive lag months and then compute age-to-age factors as the ratio of those averages
- Per member age-to-age ratios - divide payments per lag by exposure to create PMPM payments, and then apply the dollar-weighted approach as before.
Methods for adjusting development method reserve estimates for recent incurral months
Completion factors for the most recent months are typically too small to be credible and should be replaced using one of the following methods:
- Loss ratio method
- Projection method
- Credibility-weighted average of completion estimates with estimates based on the projection or loss ratio method. Weights are usually assigned based on how close the completion factor is to 1.000, with the most recent month often receiving zero completion credibility