01.06.2015 Views

Actuarial Modelling of Claim Counts Risk Classification, Credibility ...

Actuarial Modelling of Claim Counts Risk Classification, Credibility ...

Actuarial Modelling of Claim Counts Risk Classification, Credibility ...

SHOW MORE
SHOW LESS

Create successful ePaper yourself

Turn your PDF publications into a flip-book with our unique Google optimized e-Paper software.

Mixed Poisson Models for <strong>Claim</strong> Numbers 29<br />

where is the annual expected claim number and d is the length <strong>of</strong> the observation period<br />

(the exposure-to-risk). The probability mass function can be expressed using the generalized<br />

binomial coefficient:<br />

( )<br />

a + k a a ( ) d k<br />

PrN = k =<br />

ak + 1 a + d a + d<br />

( )( ) a + k − 1 a a ( ) d k<br />

=<br />

k= 0 1 2<br />

k a + d a + d<br />

Henceforth, we write N ∼ ina d to indicate that N obeys the Negative Binomial<br />

distribution with parameters a and d. This model has been applied to retail purchasing,<br />

absenteeism, doctor’s consultations, amongst many others.<br />

Moments<br />

If X ∼ am , its mean is EX = / and its variance is VX = / 2 .IfN ∼<br />

ina d then the mean is EN = d and the variance is VN = d+d 2 /a according<br />

to (1.29). It can be shown that / √ V = 2, in (1.31) for the Negative Binomial<br />

distribution.<br />

Probability Generating Function<br />

If X ∼ am , its moment generating function is<br />

(<br />

Mt = 1 −<br />

) t −<br />

if t< (1.36)<br />

The probability generating function <strong>of</strong> N ∼ ina d is<br />

(<br />

)<br />

a<br />

a<br />

N z =<br />

(1.37)<br />

a − dz − 1<br />

This result comes from (1.33) together with (1.36).<br />

True and Apparent Contagion<br />

Apparent contagion arises from the recognition that sampled individuals come from a<br />

heterogeneous population in which individuals have a constant but different propensity to<br />

experience accidents. A given individual may have a high (or low) propensity for accidents<br />

but occurrence <strong>of</strong> an accident does not make it more (or less) likely that another accident will<br />

occur. However, aggregation across heterogeneous individuals may generate a misleading<br />

statistical finding which suggests that occurrence <strong>of</strong> an accident increases the probability <strong>of</strong><br />

another accident; the observed but persistent heterogeneity can be misinterpreted as due to<br />

a strong serial dependence.<br />

True contagion refers to dependence between the occurrences <strong>of</strong> successive events. The<br />

occurrence <strong>of</strong> an event, such as an accident or illness, may change the probability <strong>of</strong><br />

subsequent occurrences <strong>of</strong> similar events. True positive contagion implies that the occurrence<br />

<strong>of</strong> an event shortens the expected waiting time to the next occurrence <strong>of</strong> the event.<br />

The alleged phenomenon <strong>of</strong> accident proneness can be interpreted in terms <strong>of</strong> true<br />

contagion as suggesting that an individual who has experienced an accident is more likely

Hooray! Your file is uploaded and ready to be published.

Saved successfully!

Ooh no, something went wrong!