Page 222 - Contributed Paper Session (CPS) - Volume 1

CPS1283 Kelvin H.C.Y et al.
(i) the new products and features introduced are appropriately
priced using suitable techniques and reliable data;
(ii) the risks associated with the development and pricing of new
products are well understood and properly mitigated, which
includes robust assessments of the potential impact of product
and pricing decisions on profitability and capital adequacy; and
(iii) adequate systems and processes are in place to support risk
analysis, pricing, and reserving at a sufficiently granular level.
Generalised Linear Models is widely accepted as the industry standard for
pricing private Motor insurance in many developed markets. In line with these
developed markets, the Malaysian actuaries widely adopted Generalised
Linear Models as their individual risk Motor pricing model during the phased
liberalisation period.

2. Methodology
As implied by its name, the Generalised Linear Model is a generalisation of
the ordinary linear regression. The linear regression is generalised by allowing
the linear model to be linked to the response variable through a link function.
Anderson et al. (2007) summarizes the structure of a Generalised Linear
Model as follow:-

= Ε[ ] = −1 (∑ +

where
is the vector of responses

g(x) is the link function: a specified (invertible) function which relates
the expected response to the linear combination of observed
factors
is a matrix produced from the factors

is a vector of model parameters, which is to be estimated

is a vector of known effects

To determine the Motor insurance premium using Generalised Linear
Models, the expected claims cost is computed based on a set of rating factors
together with its vector of model parameters. Each individual risk profile is
reflected through the rating factors, which in turn, translates into an estimated
claims cost that the general insurance company is expected to incur.
Broadly speaking, the actuary determines the rating factors to be
considered within the Generalised Linear Model. In determining the number
of rating factors to be included, the actuary needs to consider the practicality
of including such rating factors against the availability of data, customer
expectations, and permitted regulations. Based on the selected rating factors,

211 | I S I W S C 2 0 1 9