LECTURE NOTES ON:

TOPIC1: STOCHASTIC PROCESSES

STOCHASTIC MODELING FOR ACTUARIAL APPLICATION

SHAMSHIMAH SAMSUDDIN
MARCH 2021
1. Stochastic Processes

 A model that recognises the random nature of the input components is stochastic
model.
 A model that does not contain any random component is deterministic model in
nature.
 In a deterministic model, the output is determined once the set of fixed inputs and the
relationships between them have been defined. By contrast, in a stochastic model the
output is random in nature – like the inputs, which are random variables. The output is
only a snapshot or an estimate of the characteristics of the model for a given set of
inputs.
 Several independents runs are required for each set of inputs so that statistical theory
can be used to help in the study of the implications of the set of inputs.
 A deterministic model is really just a special (simplified) case of a stochastic model.
 Models may also be constructed in discrete or continuous state spaces or with discrete
or continuous time.
 A stochastic process is a family or set of ordered random variables.
 The order is indicated by indexing each random variable in the family by subscript.
 Generally, the ordering is a result of the random variables being observed over time,
so X t is a random variable that models the value of the stochastic process at time t .
 The random variables in the set can be dependent on one another reflecting the nature
of the process being modelled.
 For example, we may be concerned with the capital of an insurance company. This
fluctuates in a random manner with time since, the number of policies the company
has on its books, and hence its income, varies at random with time, the number and
the times of claims is random and so is the size of the individual claims. Therefore, say
that the capital of the company is a random or stochastic process. X t might model
the size of the individual claims at time t , and the observations of the size of claims for
the last 365 days. So, we can use these data to describe the process and to analyse
the nature of its past behaviour over time. The data may also be used to estimate the
parameters of our stochastic process model. We could use then use the estimated
stochastic process model to predict the future behaviour of the size of individual
claims. It is the dependence between the random variables in the set that allows us to
make predictions by extrapolating past patterns into the future.
now

time
X0 X1 X2 X3 X20 X18 X19 X20 Model

x0 x1 x2 x3 Sample

observe predict

describe & analyse

 Definition
A stochastic process is a model for a time-dependent random phenomenon.
Single random variable describes a static random phenomenon while stochastic
process is a collection of random variables X t , one for each time t in some set J ,
denoted X t : t  J  .
The set of value that the random variables X t are capable of taking is called state
space of the process, S . Just like a sample point is an outcome of a random variable,
so a sample path is an outcome of a stochastic process.
One should keep in mind that in the background there is some sample space  on
which all the random variables X t , t  J , are defined.
So, a sample path of the stochastic process is nothing more than the function
t  X t   for some fixed    .
Note that in what follows we shall usually suppress the dependence of X t on  in the
notation.
 Sample path- a joint realisation of the random variables X t for all t in J .
 The value in the set J are called the time set of the process.
 The set of random variables may be dependent. So, in order to describe their
statistical properties, we will need to know their joint distribution.
 If the random variables in the set were independent (as is the case for example with
statistical models of the sampling process) it would be sufficient to know the (marginal)
distribution of each random variable in the set.
 The random variables in the set need not be identically distributed. However, we will
discover that processes that have identically distributed random variables are
particularly important in the study of stochastic processes.
 The state space of the stochastic process will include all of the values that can be
taken by any of the random variables in the set. However, for particular random
variables in the set, some of these values may have zero probability of occurring.
 For example, we might model the size of individual claims by a stochastic process
X t . The random variable X t models the value at the end of day t . The value of X t
might depends on the values at the end of the two previous claim days X t 1 and X t  2 . If
this dependence does not change with t , then we could use the model to predict
future values of X t . Note that these predictions may not be exact even if the model is
good.
 So, selecting a stochastic process to model a real life situation is that of the nature
(discrete or continuous) of the time set J and of the state space S .
 The decision to use a discrete or continuous state model for a particular system is
driven by the objectives of the study, rather than whether or not the system itself is of a
discrete or continuous nature. A model may also consider time in a discrete or
continuous way. This may reflect the fact that outputs from the model are only required
at discrete points in time, or may be to satisfy the objectives of the modelling.
 Time set – the time set for this stochastic process is J , which contains all points at
which the value of the process can be observed. The time set can be discrete or
continuous.
 State space– the state space of stochastic process X t : t  J  is the set of values that
the random variable X t can take. The state space can be discrete or continuous.
 Examples:
Discrete State Continuous State
Discrete Time A motor insurance company Customers enter a queue at
reviews the status of its random i.e times between
customers’ No Claim Bonus successive arrivals are random,
(NCB) yearly. Under this State space- 0 ,  
company’s NCB scheme three Time set- 1,2,…
levels of discount are possible
0%, 25% and 40% depending on
the accident record of the driver.
State space- 0,25,40
Time set- 0,1,2,…
Continuous An insurance company is An insurance company is
Time interested in the number of claims interested in the total value of the
that is receives by time. amount of claims that it receives
State space- 0,1,2,… by time.
Time set- 0 ,   State space- 0 ,  
Time set- 0 ,  
 Processes of mixed type
Just because a stochastic process operates in continuous time does not mean that it
cannot also change value at predetermined discrete instants such processes are said
to be of mixed type.
For example: Pension scheme in which members have the option to retire on any
birthday between ages 55 to 60. The number of people electing to take retirement at
each year of age between 55 to 60 cannot be predicted exactly, nor can the time and
number of deaths among active members. Hence, the number of contributors to the
pension scheme can be modelled as a stochastic process of mixed type with state
space S  0 ,1,2 ,3,... and time set J  0 ,   . Decrements of random amounts will occur
at fixed dates due to retirements as well as at random dates due to death.
So, this process is a mixture of two processes:
 A stochastic process modelling the number of deaths which has discrete state
space and a continuous time set.
 A stochastic process modelling the number of early retirements which has a
discrete state space and a discrete time set.
 It is important to be able to conceptualise the nature of the state space of any process
which is to be analysed, and to establish whether it is most usefully modelled using a
discrete, a continuous, or a mixed time set.
 Usually the choice of state space will be clear from the nature of the process being
studied, but whether a continuous or discrete time set is used will be often depend on
the specific aspects of the process which are of interest, and upon practical issues.
 Counting process - A counting process is a stochastic process, X, in discrete or
continuous time, whose state space S is the collection of natural numbers {0, 1, 2, …},
with the property that X(t) is a non-decreasing function of t.