CT4: Q&A Bank Part 1 – Questions Page 1

Part 1 – Questions

Question 1.1

Explain what is meant by “a stochastic model” and give two advantages these have over
deterministic models. [3]

Question 1.2

For a discrete time stochastic process X n define the terms:

(i) stationary [1]

(ii) weakly stationary [1]

(iii) increment [1]

(iv) Markov property [1]

[Total 4]

Question 1.3

A moving average (stochastic) process, X n , has a discrete time domain and a

continuous state space and is defined as:

X n = Z n + a1Z n-1 + a 2 Z n- 2 + a 3 Z n-3

( )
where [ Z n ]n =-• are independent and identically distributed N 0, s 2 random variables
n =+•

and a1, a 2 , a 3 are constants.

(i) Prove that X n is weakly stationary. [5]

(ii) Explain whether the Markov property holds. [2]

(iii) Deduce whether the process has independent increments. [1]

[Total 8]

The Actuarial Education Company © IFE: 2005 Examinations

Page 2 CT4: Q&A Bank Part 1 – Questions

Question 1.4

(i) (a) Define a Poisson process with rate l .

(b) Define a compound Poisson process. [2]

(ii) The cumulative amount of claims reaching an insurance company is modelled

using a compound Poisson process.

(a) Prove that the compound Poisson process has the Markov property.

(b) Does this seem reasonable for the given insurance model?

(c) Is the process weakly stationary?

(d) Would you expect the cumulative insurance claims to follow a weakly
stationary process? [5]
[Total 7]

Question 1.5

Consider a homogeneous Markov chain process { X t }t•=0 on state space S = {0,1, 2,3}
with transition matrix:

Ê 110 2
10 0 7
10ˆ
Á 0 1 1 0˜
P=Á2 2 2 ˜
Á 5 0 2
5
1 ˜
5
Á1 9 ˜
Ë 10 0 0 10¯

(i) (
Calculate P X 5 = 3 X 4 = 2 . ) [1]

(ii) (
Calculate P X t + 2 = 1 X t = 2 . ) [2]
[Total 3]

© IFE: 2005 Examinations The Actuarial Education Company

CT4: Q&A Bank Part 1 – Questions Page 3

Question 1.6

A simple NCD system for motor insurance has four levels of discount – 0%, 20%, 40%
and 60%. A new policyholder starts on 0% discount. At the end of each policy year,
policyholders will change levels according to the following rules:
● At the end of a claim free year, a policyholder moves up one level, or remains on
the maximum discount.
● At the end of a year in which exactly one claim was made, a policyholder drops
back one level, or remains at 0%.
● At the end of a year in which more than one claim was made, a policyholder
drops back to zero discount.

For a particular driver in any year, the probability of a claim free year is 0.9, the
probability of exactly one claim is 0.075, and the probability of more than one claim is
0.025.

A policyholder takes out a policy for the first time on 1 January 2001, and by
1 January 2004 he/she had made only one claim, on 3 May 2003. What is the
probability that he/she is on 20% discount in 2006? [3]

Question 1.7

A Markov chain is determined by the transition matrix

Ê 13 1
6
1
2 0ˆ
Á1 0 1 0˜
P=Á 2 2 ˜
Á 16 1
6
2
3 0˜
Á0 1˜¯
Ë 0 0

Which of the following are stationary distributions for the chain?

I ( 9 35 , 17 , 3 5 , 110)
II (0, 0, 0,1)

III ( 9 70 , 114 , 310 , 1 2 ) [3]

The Actuarial Education Company © IFE: 2005 Examinations

Page 4 CT4: Q&A Bank Part 1 – Questions

Question 1.8

A Markov chain { X n }•n=0 has a discrete state space S. The initial probability
distribution is given by P [ X 0 = i ] = qi . The one-step transition probabilities are

denoted by P [ X m +1 = im +1 | X m = im ] = pi( i
m,m +1)
.
m m +1

(i) State the Markov property for such a process. [1]

(ii) Write down expressions for the following in terms of p ’s and q ’s.

(a) P [ X 0 = i0 , X1 = i1,… , X n = in ]

(b) P [ X 4 = i] [2]
[Total 3]

Question 1.9

A new actuarial student is planning to sit one exam each session. He expects that his
performance in any exam will only be dependent on whether he passed or failed the last
exam he sat. If he passes a given exam, the probability of passing the next will be a ,
regardless of the nature of the exam. If he fails an exam, the probability of passing the
next will be b .

(i) What is the probability that:

(a) the first exam he fails is the seventh, given that he passes the first?

(b) he passes the fifth exam, given that he fails the first three? [4]

(ii) Explain the results above in terms of a Markov chain, specifying the state space
and transition matrix. (For the purposes of this model, assume that we are only
interested in predicting passing or failing, not in the number of exams passed so
far.) [3]
[Total 7]

© IFE: 2005 Examinations The Actuarial Education Company

CT4: Q&A Bank Part 1 – Questions Page 5

Question 1.10

An insurance company operates a no claims discount system with discount levels of 0%,
30%, 40%, 50% and 60%. The rules are as follows:
● At the end of a claim free year, a policyholder moves up one level (or remains
on maximum discount).
● At the end of a year in which exactly one claim was made, a policyholder drops
back two levels (or moves to zero discount).
● At the end of a year in which more than one claim was made, a policyholder
drops back to zero discount.

What premium will a motorist who first took out a policy on 1 January 1993 be paying
for insurance cover in 2004, if the policyholder made claims on 15 August 1994,
3 February 1998, 17 September 1998 and 14 November 2003, and the full premium for
2004 is £750 pa? [4]

Question 1.11

Determine all the stationary distributions for a Markov chain with transition matrix:

Ê0 0 0 1
2
1
2ˆ
Á0 1 4 0 0˜
Á 5 5 ˜
P=Á 0 0 0 1
3
2 ˜
3
Á0 4 1 0 0 ˜˜
Á 5 5
ÁË 1 3 1 0 0 ˜¯
2 10 5
[5]

The Actuarial Education Company © IFE: 2005 Examinations

Page 6 CT4: Q&A Bank Part 1 – Questions

Question 1.12

Quiz show contestants must answer questions in order to win a one million pound
jackpot. Let pi be the probability that they win the jackpot when they currently have
£100,000i. They receive £100,000 for a correct answer, but they lose £100,000 for a
wrong answer. The game is played until they are broke or they win the jackpot. The
probability that they get a question correct is 2 3 independently of any other questions
or answers.

(i) Set up a difference equation and two boundary conditions for pi , and show that

1 - ( 12 )
i
pi =
1 - ( 12)
10

is a solution. [4]

(ii) What is the probability that they win the jackpot given that they start with
£500,000? [2]
[Total 6]

© IFE: 2005 Examinations The Actuarial Education Company

CT4: Q&A Bank Part 1 – Questions Page 7

Question 1.13

The stochastic process { X t } is defined by the relationship X t = Z t + Zt -1 , where {Zt }

is a sequence of uncorrelated random variables with probability function:

Ï1 with probability p
Zt = Ì
Ó1, 000 with probability q

where p + q = 1 and q < p .

(i) Find expressions in terms of p and q for each of the following probabilities:

(a) P( X 5 = 1, 001)

(b) P( X 5 = 1, 001 | X 4 = 1, 001)

(c) P( X 5 = 1, 001 | X 4 = 1, 001, X 3 = 1, 001) . [6]

(ii) State, with reasons, whether { X t } has the Markov property. [2]
[Total 8]

The Actuarial Education Company © IFE: 2005 Examinations

Page 8 CT4: Q&A Bank Part 1 – Questions

Question 1.14

At the end of each year an independent organisation ranks the performance of the unit
trusts invested in a particular sector, and classifies them into four quartiles (with
quartile 1 being the best performer). Past experience has shown that, at the end of each
year, a fund will either remain in the same quartile or will move to a neighbouring
quartile.

In fact, there is a probability 1 - 2a that a fund will remain in the same quartile and,
where upward or downward movements are both possible, these are equally likely.
However, it has been found that a fund that has remained in the top or bottom quartile for
two consecutive years has a probability of 1 - b ( b < a ) of remaining in the same
quartile the following year.

(i) Construct a discrete-time Markov chain with six states to model this situation,
defining the states in your model and drawing a state transition diagram. [3]

(ii) Write down the transition matrix for your model. [2]

(iii) Explain whether this Markov chain is irreducible, periodic or both. [2]

(iv) Show that, if a stationary distribution exists with a quarter of the funds in each
a (1 - 2a )
quartile, then b = . [4]
1-a

(v) Last year 20% of funds in the second quartile moved up to the top quartile.
Assuming the fund rankings have reached a stationary state, estimate the
proportion of funds in the top quartile last year that have remained in the top
quartile this year. [2]
[Total 13]

© IFE: 2005 Examinations The Actuarial Education Company

CT4: Q&A Bank Part 1 – Questions Page 9

Question 1.15

A time-homogeneous Markov chain X t with states labelled 1,2,3,4,5, is defined by the

following transition matrix:

Ê 0.2 0.3 0.2 0.2 0.1ˆ

Á 0 0.3 0 0.4 0.3˜
Á ˜
P = Á 0.2 0 0.7 0 0.1˜
Á 0 0 0 0.8 0.2˜
Á ˜
ÁË 0 0.3 0 0.7 0 ˜¯

Calculate the following:

(i) P ( X10 = 3 | X 8 = 3) [1]

(ii) P ( X 4 = 3 | X 0 = 4) [1]

(iii) lim P ( X n = 4 | X 3 = 1) [4]

nƕ

(iv) P ( R10 = 7 | X10 = 4) , where Rt is the residual holding time at time t, ie the
length of time you have to wait before leaving state 4. [1]
[Total 7]

The Actuarial Education Company © IFE: 2005 Examinations

Page 10 CT4: Q&A Bank Part 1 – Questions

Question 1.16

A gambler plays roulette in a casino, always betting £1,000 on red. This gives a
probability p = 18 37 < ½ of winning on any particular turn, since 0 is white. If they win
they receive £1,000 on top of their stake. If they lose they forfeit their stake. They will
play either until they are broke, or until they have made £50,000. They start with
£10,000.

Let pi denote the probability of ultimate ruin, when the gambler’s current wealth is i
(in units of £1, 000 ).

(i) Construct a difference equation and boundary conditions satisfied by pi . [2]

( )
i
(ii) Show that pi = c1 + c2 q
p satisfies the equation, where q = 1 - p and c1 and
c2 are constants. [2]

(iii) By using the boundary conditions, determine c1 and c2 . [2]

(iv) Hence calculate the probability of ultimate ruin. [2]

[Total 8]

© IFE: 2005 Examinations The Actuarial Education Company

CT4: Q&A Bank Part 1 – Questions Page 11

Question 1.17

A simple NCD system has four levels of discount – 0%, 20%, 40% and 60%. A new
policyholder starts on 0% discount. At the end of each policy year, policyholders will
change levels according to the following rules:
● At the end of a claim free year, a policyholder moves up one level, or remains on
the maximum discount.
● At the end of a year in which exactly one claim was made, a policyholder drops
back one level, or remains at 0%.
● At the end of a year in which more than one claim was made, a policyholder
drops back to zero discount.

7
For a particular policyholder in any year, the probability of a claim free year is 10 , the
probability of exactly one claim is 1 and the probability of more than one claim is 1 .
5 10

(i) Write down the transition matrix for this time homogeneous Markov chain. [2]

Calculate the 2-step transition probabilities from state i to state j , pij( ) .

2
(ii) [3]

(iii) If the policyholder starts with no discount, what is the probability that this
policyholder is at the maximum discount level 5 years later? [5]

(iv) If a large number of people having the same claim probabilities take out policies
at the same time, what proportion would you expect to be in each discount
category after a long time? [5]
[Total 15]

The Actuarial Education Company © IFE: 2005 Examinations

Page 12 CT4: Q&A Bank Part 1 – Questions

Question 1.18

Show that, under the assumptions of the two-state Markov model for mortality:

var[ Di - mVi ] = E[ Di ] [6]

The notation used here is the same as in the Core Reading, and you may assume that
E[ Di - mVi ] = 0 (which is proved in Chapter 4 of the Course Notes).

Question 1.19

A survival model for an elderly population has two states A and D, representing alive
and dead. The force of mortality at age t years m (t ) , ie the transition rate from A to D,
is given by:

m (t ) = 0.0001 ¥ (1.10)
t

Calculate the probability that a 60-year-old will survive to age 80. [2]

© IFE: 2005 Examinations The Actuarial Education Company