FCAS exams

Exam 7

Advanced Estimation of Claims Liabilities

5 discussions

Study strategy

Exam 7: Scenarios and probability distributions should not be mixed silently

A sensitivity asks how a result changes when an assumption changes. A stress is a specified adverse scenario. A predictive distribution assigns probabilities under a model. These outputs can complement one another, but they answer different questions.

Worked example or practice scenario

A +10% severity scenario raises a toy unpaid estimate from 5m to 5.5m. Without a probability model, the result is not “the 99th percentile.” Assigning a percentile label merely because a scenario seems severe invents statistical meaning.

Try this next

For each output in a practice problem, name its interpretation and assumptions. Use the current content outline and released resources to check required methods. Avoid importing an old topic list or using a scenario result as a stochastic confidence statement.

Reading sources

Official syllabus and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

0 repliesOct 1

Explanation

Exam 7: Bootstrap output depends on the resampling model

Bootstrapping is a procedure, not a guarantee that all uncertainty has been captured. The resampled units, fitted structure, residual treatment and process simulation determine what the resulting distribution represents.

Worked example or practice scenario

In an original reserving exercise, resampling residuals around a fitted development pattern explores variation under that pattern. If the real process has a changing calendar trend absent from the fit, the bootstrap does not automatically represent that structural change.

Try this next

Write the algorithm in steps: fit, calculate residuals, resample, refit or project, and include the intended process variation. Identify assumptions about dependence and heterogeneity. Explain a separate stress scenario for a limitation outside the resampling model.

Reading sources

Official syllabus and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

0 repliesOct 1

Explanation

Exam 7: A tail factor can dominate an immature estimate

A tail factor extends development beyond the observed ages. Its impact depends on the remaining amount and the uncertainty in extrapolation. Treat tail selection as an assumption with evidence, not as a default number copied between lines.

Worked example or practice scenario

If a pre-tail ultimate estimate is 8m, a tail factor of 1.05 produces 8.4m; 1.10 produces 8.8m. The 400,000 difference is material even though the factors look close. These are original sensitivity numbers, not an estimated industry tail.

Try this next

Compare relevant mature experience and alternative extrapolations, then explain applicability to the segment. Separate a tail sensitivity from a probabilistic uncertainty interval. The latter requires a specified model and distribution.

Reading sources

Official syllabus and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

0 repliesOct 1

Explanation

Exam 7: Residuals help diagnose a stochastic development model

Residuals can reveal patterns left unexplained by a fitted model. Look across origin, development and calendar periods. A pattern is a diagnostic prompt; it should lead to a data or assumption investigation rather than an automatic model adjustment.

Worked example or practice scenario

Positive residuals concentrated along a recent calendar diagonal might suggest inflation or operational change. A development-age pattern might indicate that the selected mean or variance structure fits some maturities poorly. Neither interpretation is established by the plot alone.

Try this next

State a competing explanation and a check that could distinguish it. Compare claim counts, average size and case reserves where relevant. Document whether extreme observations are legitimate losses or data errors before altering a fit.

Reading sources

Official syllabus and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

0 repliesOct 1

Explanation

Exam 7: A reserve point estimate and its uncertainty are separate outputs

A reserving method can estimate an expected ultimate while a stochastic framework addresses variability. Process uncertainty, parameter uncertainty and model limitations should be distinguished. A precise-looking estimate does not imply narrow uncertainty.

Worked example or practice scenario

Two toy portfolios can have the same best estimate of unpaid claims, 10m, but different volatility because of claim sizes, tail length or data volume. A percentile or standard error must come from the stated model and assumptions; it is not determined by the point estimate alone.

Try this next

Explain which uncertainties an assigned model includes and which it omits. Compare one-year development uncertainty with ultimate uncertainty before using a result in a capital or reporting question.

Reading sources

Official syllabus and resources

ActNet editorial guide · October 1, 2026 · Original illustrative examples.

0 repliesOct 1