SOA Exam ASTAM Sample Question 23
Study guide: Exam ASTAM
In a Bühlmann credibility model, you are given that for a portfolio of short-term insurance contracts, the number of claims X_1, \ldots, X_T, conditional on \Theta, are independent and identically distributed with
hypothetical mean: \mu(\Theta) = E\left[X_t \mid \Theta\right]
and
process variance: \nu(\Theta) = Var\left[X_t \mid \Theta\right]
Denote the structural parameters by \mu = pure premium, v = expected process variance, and a = variance of the hypothetical mean.
Part a
Prove that the unconditional variance of X_t is the sum of the expected process variance and the variance of the hypothetical mean, that is,
Var\left[X_t\right] = \nu + a.
Part b
In the context above, give an expression for the Bühlmann credibility factor Z for the number of claims for a policy observed for T years.
Part c
An insurance company analyzes its portfolio of insurance policies using the Bühlmann credibility model to estimate claim frequency. It assumes that claim frequency for a particular policy, conditional on the mean parameter \lambda, is Poisson.
Show that the following equations are true:
(i) \mu = E[\lambda]
(ii) Var\left[X_t\right] = E[\lambda] + Var[\lambda]
(iii) a = Var\left[X_t\right] - E\left[X_t\right]
Part d
You are given the following claim frequencies for the insurance company's portfolio with one year experience for each policyholder:
| Number of claims | Number of policyholders |
|---|---|
| 0 | 8,416 |
| 1 | 1,427 |
| 2 | 153 |
| 3 | 3 |
| 4 | 1 |
| Total | 10,000 |
Calculate estimates for the structural parameters \mu, \nu, and a.
Part e
Calculate the estimated Bühlmann credibility factor Z. Use it to calculate the estimated credibility premium for the number of claims for each policyholder.
Part f
Explain why only one year's experience per policyholder is needed to estimate Bühlmann credibility factors.
Answer this question and see the worked solution
Concepts tested
- Bühlmann Credibility
- Empirical Bayes Credibility
- Expected Value of Process Variance
- Variance of Hypothetical Means
- Poisson Distribution