INSTITUTE AND FACULTY OF ACTUARIES

EXAMINATION
28 April 2011 (am)

Subject CT6 Statistical Methods Core Technical

Time allowed: Three hours INSTRUCTIONS TO THE CANDIDATE 1. Enter all the candidate and examination details as requested on the front of your answer booklet. You must not start writing your answers in the booklet until instructed to do so by the supervisor. Mark allocations are shown in brackets. Attempt all 11 questions, beginning your answer to each question on a separate sheet. Candidates should show calculations where this is appropriate.

2.

3. 4. 5.

Graph paper is NOT required for this paper.

AT THE END OF THE EXAMINATION Hand in BOTH your answer booklet, with any additional sheets firmly attached, and this question paper.
In addition to this paper you should have available the 2002 edition of the Formulae and Tables and your own electronic calculator from the approved list.

CT6 A2011

Institute and Faculty of Actuaries

Give two examples of exercises where Monte-Carlo simulation should be performed using the same choice of random numbers, explaining your reasoning in each case. [4]

An insurance company has collected data for the number of claims arising from certain risks over the last 10 years. The number of claims in the jth year from the ith risk is denoted by Xij for i = 1, 2, , n and j = 1, 2, , 10. The distribution of Xij for j = 1 2, , 10 depends on an unknown parameter i and given i the Xij are independent identically distributed random variables. (i) Give a brief interpretation of E[s2()] and V[m())] under the assumptions of Empirical Bayes Credibility Theory Model 1. [2] Explain how the value of the credibility factor Z depends on E[s2()] and V[m()]. [3] [Total 5]

(ii)

Let y1, , yn be samples from a uniform distribution on the interval [0, ] where > 0 is an unknown constant. Prior beliefs about are given by a distribution with density

(1+ ) f () = 0

> otherwise

where and are positive constants. (i) Show that the posterior distribution of given y1 is of the same form as the prior distribution, specifying the parameters involved. [4] Write down the posterior distribution of given y1, , yn. [2] [Total 6]

(ii)

The annual number of claims on an insurance policy within a certain portfolio follows a Poisson distribution with mean . The parameter varies from policy to policy and can be considered as a random variable that follows an exponential distribution with mean 1 . Find the unconditional distribution of the annual number of claims on a randomly chosen policy from the portfolio. [6]

CT6 A20112

The number of claims under an insurance policy in a year is either 0 (with probability 40%) or 1 (with probability 20%) or 2 (with probability 40%). Individual claim amounts are equally likely to be 50 or 20. The insurance company calculates premiums using a premium loading of 50% and is considering operating one of the following arrangements: (A) (B) Making no changes. Introducing a policy excess of 10 (per claim) in return for a reduction of 5 in premiums. Effecting an individual excess of loss reinsurance arrangement with retention 30 for a premium of 10.

(C)

Construct a table of the insurance companys profits under all the possible outcomes for each of (A) (B) and (C) and hence determine the optimal arrangement using the Bayes criteria. [8] The double exponential distribution with parameter > 0 has density given by g(x) = e|x| x . (i) (ii) Construct an algorithm for generating samples from this distribution. [3]

Construct an algorithm for producing samples from a N(0,1) distribution using samples from the double exponential distribution and the acceptance-rejection method. [6] [Total 9]

Consider the time series Yt = 0.7 + 0.4Yt1 + 0.12Yt2 + et where et is a white noise process with variance 2. (i) (ii) (iii) (iv) Identify the model as an ARIMA(p,d,q) process. Determine whether Yt is a stationary process. Calculate E(Yt). Calculate the auto-correlations 1, 2, 3 and 4. [1] [2] [2] [4] [Total 9]

CT6 A20113

PLEASE TURN OVER

Suppose that Y is a random variable belonging to a special subset of the exponential family where the density function of Y has the form y b() f ( y, , ) = exp + c( y, ) For some constants and and functions b and c. (i) Show that the moment generating function of Y is given by b( + t ) b() M Y (t ) = exp

[3]

Hint: Note that the function f(y, + t, ) is the density of another random variable of the same family and hence (ii)

f ( y, + t , )dy = 1.
[4] [4] [Total 11]

Show that E(Y) = b() and Var(Y) = b() using the result in (i).

(iii)

Verify that the result in (i) holds if Y has a Poisson distribution.

Claims on a portfolio of insurance policies arise as a Poisson process with parameter . Individual claim amounts are taken from a distribution X and we define mi = E(Xi) for i = 1, 2, . The insurance company calculates premiums using a premium loading of . (i) (ii) Define the adjustment coefficient R. (a) Show that R can be approximated by
expansion of MX(t). (b) Show that there is another approximation to R which is a solution of the equation m3y2 + 3m2y 6m1 = 0. [6] 2m1 by truncating the series m2

[1]

Now suppose that X has an exponential distribution with mean 10 and that = 0.3. (iii) Calculate the approximations to R in (ii) and (iii) and compare them to the true value of R. [6] [Total 13]

CT6 A20114

10

The number of claims on a portfolio of insurance policies has a Poisson distribution with mean 200. Individual claim amounts are exponentially distributed with mean 40. The insurance company calculates premiums using a premium loading of 40% and is considering entering into one of the following re-insurance arrangements: (A) (B) No reinsurance. Individual excess of loss insurance with retention 60 with a reinsurance company that calculates premiums using a premium loading of 55%. Proportional reinsurance with retention 75% with a reinsurance company that calculates premiums using a premium loading of 45%. Find the insurance companys expected profit under each arrangement. [6]

(C)

(i) (ii)

Find the probability that the insurer makes a profit of less than 2000 under each of the arrangements using a normal approximation. [8] [Total 14]

11

The table below shows cumulative claims paid on a portfolio of insurance policies. Development Year 1 2 281.4 302 320 322 312.9

Accident Year 2007 2008 2009 2010

0 240 260 270 276

3 305

All claims are fully run off by the end of development year 3. (i) Calculate the total reserve for outstanding claims using the basic chain ladder technique. [7]

An actuary is considering modelling the future claims assuming that individual development factors are lognormally distributed with the following parameters: Development Year 1 to 2 0.035850 0.045606

Parameter (ii)

0 to 1 0.171251 0.032148

2 to 3 0.008787 0.046853

Show that under these assumptions the cumulative development factor to ultimate is also lognormally distributed.

[3]

(iii)

Calculate a 99% upper confidence limit for the outstanding claims relating to the 2010 accident year. [5] [Total 15]

END OF PAPER

CT6 A20115