Depending on the detail of the data, and availability of historical data, a variety of actuarial methods may be used to determine ultimate loss, including:
- Paid Loss Development
- Incurred Loss Development
- Paid Bornhuetter-Ferguson Methodology
- Incurred Bornhuetter-Ferguson Methodology
- Frequency-Severity Methodology
- Backward Recursive Technique
- Berquist-Sherman Technique
- Cape Cod Technique
These methods capture both adverse developments on known cases (incurred but not enough reported, or IBNR) as well as anticipated late emergence of claims (incurred but not reported, or IBNR).
All of the methods produce estimates of ultimate loss as of a specific date for historical policy/accident years. Estimates of outstanding losses are derived by subtracting paid losses to date from selected ultimate losses. Discount factors will be derived using expected payout patterns and applied to undiscounted reserves to calculate discounted reserves. For each line, we will develop ranges of ultimate and outstanding losses, as well as mid-range estimates. Adjusted IBNR levels as of the report date will be based on the reporting pattern used to derive the ultimate losses.
Huggins uses a frequency/severity Monte Carlo simulation in order to calculate the confidence intervals around our central estimates of reserves. For each line of business, we will calculate the frequency and severity components of the indicated reserves by accident year using claim count information. The frequency component will be the number of open plus unreported claims by accident year. The severity component by accident year will be calculated as the central estimate of reserves divided by open plus unreported claims, which will show the average claim size for open/unreported claims. The frequency and severity components of the reserves by accident year will be used as inputs in the Monte Carlo simulation performed to estimate the variability around the central estimate of reserves. We derive confidence intervals by simulating the expected results over and over for an extremely large number of trials and recording the results of each simulation.