Mid Test

April 2010

Q1.

(a) Let 𝑋𝑗 be the claim experience of the 𝑗 𝑡ℎ policy in a group, where 𝑗 ∈ {1,2, … , 𝑛}. It is assumed
1
that the 𝑋𝑗 ′𝑠 are independent. Suppose 𝐸[𝑋𝑗 ] = 𝜉 and 𝑉𝑎𝑟[𝑋𝑗 ] = 𝜎 2 . If 𝑋̅ = ∑𝑛𝑖=1 𝑋𝑖 , assign full
𝑛
credibility if 𝑃𝑟[−𝑟𝜉 ≪ 𝑋̅ − 𝜉 ≪ 𝑟𝜉] ≥ 𝑝 for a particular value of 𝑟 > 0 and 𝑝 > 0. If 𝑛 is large,
determine the standard for full credibility in terms if the number of exposure units. (Show the
work)
𝑁 𝑗
(b) Assume further that 𝑋𝑗 is compound Poisson distributed, that is 𝑋𝑗 = ∑𝑖=1 𝑌𝑗 , where each 𝑁𝑗 is
Poisson with parameter 𝜆 and the claim size distribution 𝑌 has mean 𝜃𝑌 and variance 𝜎𝑌2 . If the
accuracy is to be measured with regards to the average of total claims the determine the standard
full credibility in terms of the number of policies, in terms of the expected number of claims, and
in terms of the expected total dollars of claims. (Show the work)

Q2.

The full-credibility standard is set so that the total number of claims is within 5% of the true value
with probability 𝑝. This standard is 800 claims. The standard is then altered so that the total cost of
claims is to be within 10% of the true value with probability 𝑝. The number of claims has a Poisson
distribution; and the claim severity follows a Lognormal distribution with 𝜇 = 7 and 𝜎 = 2. Determine
the expected number of claims necessary to obtain full credibility under the new standard.

Q3.

The following information is given:

1. The number of claims for a single insured follows a Poisson distribution with mean M
2. The amount of a single claim has an exponential distribution with pdf 𝑓𝑋|Λ (𝑥|𝜆) = 𝜆−1 𝑒 −𝑥/𝜆 , 𝑥 >
0 and 𝜆 > 0
3. 𝑀 and Λ are independent
4. 𝐸[𝑀] = 0.10 and 𝑉𝑎𝑟[𝑀] = 0.0025
5. 𝐸[Λ] = 1,000 and 𝑉𝑎𝑟[Λ] = 640,000
6. The number of claims and the claim amounts are independent

A randomly selected risk has the following experience over three years:

Year Number of Exposures Aggregate Losses

1 30 35,000
2 24 30,000
3 26 24,000
Calculate the B𝑢̈ hlmann-Straub estimate of the expected aggregate loss per exposure in Year 4 for
this risk.

Q4. For a portfolio of insurance policies, the annual claim amount of 𝑋 of a policy has the following
probability density function:
 2𝑥
𝑓𝑋|Θ (𝑥|𝜃) = ,0 < 𝑥 < 𝜃
𝜃2
The prior distribution of Θ has the following probability density function:

𝑓Θ (𝜃) = 4𝜃 3 , 0 < 𝜃 < 1

A randomly selected policy has claim amount 0.1 in Year 1. Determine the B𝑢̈ hlmann credibility
estimate of the expected claim amount of the selected policy in Year 2.

March 2011

Q1.

(a) sd April 2010

(b) sd April 2010
𝜎 2
(c) Let say that 𝑋̅ is approximately normally distributed with mean 𝜉 and variance 𝑛 . Let say that full
credibility is inappropriate and that the credibility premium, 𝑃𝑐 , is given by the equation 𝑃𝑐 =
𝑍𝑋̅ + (1 − 𝑍)𝑀, where 0 ≪ 𝑍 ≪ 1 and M is the externally obtained mean. Show that 𝑍 =
𝜉 𝑛 𝑍(1+𝑝)/2 2
𝑚𝑖𝑛 {𝜎 √𝜆 , 1} where 𝜆0 = ( 2
) and 𝑍(1+𝑝)/2 is the (1 + 𝑝)/2 percentile of the standard
0

normal distribution.
Hint:
𝜉2
𝑉𝑎𝑟[𝑃𝑐 ] =
𝜆0
Q2. sd April 2010, the claim severity distribution has probability density function 𝑓(𝑥) =
0.0002(100 − 𝑥), 0 < 𝑥 < 100.
Hint: Use the results in Q1 above.

Assignment March 2012

Q1. You are given:

(i) The annual number of claims for each policyholder follows a Poisson distribution with mean
𝜆.
(ii) The distribution of 𝜆 across all policyholders has pdf:
𝑓Λ (𝜆) = 𝜆𝑒 −𝜆 , 𝜆 > 0

A randomly selected policyholder is known to have had at least one claim last year. Determine the
posterior probability that this same policyholder will have at least one claim this year.
∞ 1
Hint: ∫0 𝜆𝑒 −𝑛𝜆 𝑑𝜆 = 𝑛2

Q2. Let say that given Θ = 𝜃 the random variables 𝑋1 , 𝑋2 , … , 𝑋𝑛 are independent and identically
exponentially distributed with pdf

𝑓𝑋𝑗|Θ (𝑥𝑗 |𝜃) = 𝜃𝑒 −𝜃𝑥𝑗 , 𝑥𝑗 > 0

and Θ is itself gamma distributed with 𝛼 > 1 and 𝛽 > 0

 𝜃 𝛼−1 𝑒 −𝜃/𝛽
𝜋(𝜃) =
Γ(𝛼)𝛽 𝛼
(i) Show that the unconditional distribution of 𝑋𝑗 is a two-parameter Pareto distribution
𝛽 −1
(ii) Show that 𝐸[𝑋𝑗 ] = 𝛼−1 ,𝛼 > 1
(iii) Determine the posterior pdf 𝜋Θ|𝑋 (𝜃|𝑥)
(iv) Determine the posterior mean 𝐸[Θ|𝑥]

Hint: If a random variable Y follows a two-parameter Pareto distribution with parameters 𝛼 and 𝜃,
then the pdf of Y is
𝛼𝜃 𝛼
𝑓𝑌 (𝑦) = , 𝑦 > 0, 𝛼 > 0, 𝜃>0
(𝑦 + 𝜃)𝛼+1
Q3. Let the number of claims per automobile insured follows a Poisson distribution with mean 𝜆. The
prior distribution for 𝜆 has the following density function:

(500𝜆)50 𝑒 −500𝜆
𝑓(𝜆) =
𝜆Γ(50)
A company observes the following claims experience:

Year 1 Year 2
Number of claims 75 210
Number of automobiles insured 600 900
The company expects to insure 1100 automobiles in Year 3. Determine the expected number of claims
in Year 3.

April 2012

Q1. The Slippery Rock Insurance Company is reviewing their rates. The expected number of claims
necessary for full credibility is to be determined so that the observed total cost of claims should be
within 5% of true value 90% of the time. If 𝑋 is a random variable which denote individual claims,
based on independent studies, they have estimated that individual claims are independently and
identically distributed as following
1
𝑓(𝑥) = , 0 ≪ 𝑥 ≪ 200,000
200,000
Assume that the number of claims follows a Poisson distribution

(a) Determine the coefficient of variation of the severity distribution

(b) Derive the standard for full credibility (Show the work)
(c) What is the credibility Z to be assigned to the most recent experience given that it contains 1,082
claims? (Calculate to 3 decimal points)

Q2. For a particular policyholder, the manual premium is 600 per year. The past claims experience is
given the table below:
 Year 1 2 3
Claim amount 475 550 400
Asses whether full or partial credibility is appropriate and determine the net premium for next year’s
claims assuming the normal approximation. Use r= 0.05 and p=0.9

Q3. The number of claims for a randomly selected insured has the Poisson distribution with parameter
𝜃. The parameter 𝜃 is distributed across the population with pdf 𝜋(𝜃) = 3𝜃 −4 , 𝜃 > 1. For an
individual, the parameter does not change over time. A particular insured experienced a total of 20
claims in the previous two years. Determine the Bayesian credibility estimate for the future expected
claim frequency for this particular insured.

Q4. The amount of a claim has the exponential distribution with mean 1/𝜃. Among the class of
insureds and potential insureds, the parameters 𝜃 varies according to the gamma distribution with
𝛼 = 4 and scale parameter 𝛽 = 0.001. Suppose a person had claims of 100, 950 and 450.

(a) Determine the posterior distribution of 𝜃

(b) Determine the predictive distribution of the fourth claim
(c) Determine the Bayesian premium

March 2013

Q1. Claim frequency has a Binomial distribution with the probability of claim per policy being 0.068.
Assume full credibility is based on 98% coverage of observed claim frequency within 4% of the true
mean. Determine the credibility factor if there are 68,000 policies.

Q2. The number of claims in one year has the Binomial distribution with 𝑛 = 3 and 𝜃 unknown. The
prior distribution for Θ is beta with the probability density function 𝜋(𝜃) = 280𝜃 3 (1 − 𝜃)4 , 0 < 𝜃 <
1. There were three claims observed. Determine each of the following:

(a) The posterior distribution of Θ

(b) The expected value of Θ from the posterior distribution

Q3. An insurance company has determined that the limited-fluctuation full credibility standard is
2,000 if:

(a) the total number of claims is to be within 3% of the expected value with probability 1 − 𝛼; and
(b) the number of claims follow Poisson distribution

The standard is then changed so that the total cost of claims is to be within 5% of the expected value
with probability 1 − 𝛼, where claim severity is distributed as Uniform [0, 10000]. Determine the
expected number of claims necessary to obtain full credibility under the new standard.

Q4. Annual claim frequencies follow a Poisson distribution with mean 𝜆. The prior distribution of Λ
has probability density function
𝜆
1 1
𝑓Λ (𝜆) = 0.4 (6 𝑒 −6 ) + 0.6(12 𝑒 −𝜆/12 ), 𝜆 > 0

Ten claims are observed for an insured in Year 1. Determine the Bayesian expected number of claims
for the insured in Year 2.
March 2013

Q1. For a particular general insurance business, there are 3000 policies in a particular period. For each
policy, the claim numbers are independent and follows a negative binomial distribution with
parameters 𝑟 = 2 and 𝛽 = 0.3. The claim amounts are also independent and identically distributed
random variables following a Pareto distribution with 𝛼 = 4 and 𝜃 = 1500. The full credibility
standard is set to be within 5% the expected aggregate losses 90% of the time. Using classical
credibility theory, determine partial credibility factor of the loss experience in that period.

Q2. Let the distribution of the claims severity has mean 26. In Year 1, 1200 claims were settled with
mean severity of 32. Based on the limited-fluctuation approach, severity per policy for Year 2 was
then revised to 29.82. In year 2, 1500 claims were settled with mean severity of 24.46. what is the
revised mean severity prediction for Year 3, if the same actuarial assumptions are used as for the
prediction for Year 2?

Hint: 𝑃𝑐 = 𝑍𝑋̅ + (1 − 𝑍)𝑀, where M may be the manual rate or the overall mean of the population

Q3. An insurance company sells three types of policies with the following characteristics:

Type of Proportion of total policies Distribution of annual claim frequency

policy
A 5% Poisson with parameter 𝜆 = 0.25
B 20% Poisson with parameter 𝜆 = 0.5
C 75% Poisson with parameter 𝜆 = 1
A randomly selected policy is observed to have one claim in each of Year 1 through Year 4. Determine
the Bayes estimate of the expected number of claims of this policyholder in Year 5.

Q4. Suppose that given Θ = 𝜃 the random variables 𝑋1 , 𝑋2 , … , 𝑋𝑛 are independent with Poisson
probability
𝑥
𝜃 𝑗 𝑒 −𝜃
𝑓𝑋𝑗 |Θ (𝑥𝑗 |𝜃) = 𝑥𝑗 !
, where 𝑥𝑗 = 0,1,2, …and 𝑗 = 1,2, … , 𝑛

and let Θ has the probability density function 𝜋Θ (𝜃).

(a) Let 𝑆 = 𝑋1 +𝑋2 + ⋯ + 𝑋𝑛 . Show that the marginal probability function of S is given by the
equation
∞ (𝑛𝜃)𝑠 𝑒 −𝑛𝜃
𝑓𝑆 (𝑠) = ∫0 𝑠!
𝜋Θ (𝜃)𝑑𝜃 , where 𝑠 = 0,1,2, …
March 2014
Q1. The number of claims is distributed as Binomial Negative with parameters 𝑟 and 𝛽 = 3 and the
claims severity takes values 1, 10, and 100 with probabilities 0.5, 0.4 and 0.1, respectively. If claim
frequency, and claim severity are independent, determine the expected number of claims needed
for the observed aggregate losses to be within 10% of the expected aggregate losses with 95%
probability
Q2. A block of health insurance policies has 2,309 claims this year, with mean claim of $250 and
standard deviation of $455. If full credibility is based on 90% coverage to within 5% the true mean
claims severity, and the previous (prior) mean severity is $260, what is the updated prediction for
the mean severity next year based on the limit fluctuation approach?
Q3.
(a) Let 𝑋1 , 𝑋2 , … , 𝑋𝑛 be a random sample from a population which follows a Poisson distribution with
parameter 𝜆. It is assumed that Λ follows a gamma distribution with parameter 𝛼 and 𝜃. Show
that the posterior distribution of Λ is also gamma but with parameter
𝛼 ∗ = 𝛼 + 𝑛𝑥̅
and
1 −1 𝜃
𝜃 ∗ = [𝑛 + ] =
𝜃 𝑛𝜃 + 1
(b) A portfolio has 100 independent and identically distributed risks. The number of claims of each
risk follows Poisson distribution with parameter 𝜆. The prior distribution of Λ is gamma with
parameter 𝛼 = 4 and 𝜃 = 0.02. In year 1, the following loss experience is observed:
Number of claims Number of risks
0 90
1 7
2 2
3 1
Total 100
Determine the Bayesian expected number of claims of the portfolio in Year 2.
Hint: 𝐸[𝑋𝑛+1 |𝑥1 , 𝑥2 , … , 𝑥𝑛 ] = 𝐸[𝐸(𝑋𝑛+1 |Λ)|𝑥1 , 𝑥2 , … , 𝑥𝑛 ]

Q4. The severity of an individual claims has the exponential distribution with probability density
1
function 𝑓𝑋|Λ (𝑥|𝜆) = 𝜆 𝑒 −𝑥/𝜆 , 𝑥, 𝜆 > 0. The parameter 𝜆 has the inverse gamma distribution with
the probability density function 𝜋(𝜆) = 400𝜆−3 𝑒 −20/𝜆 .

(a) Determine the unconditional expected value of the severity

(b) Suppose two claims were observed with values 15 and 25. Determine the Bayesian credibility
estimate of the expected value of the next claim rom the same insured.