A comparison of methods for modelling rates of withdrawal from insurance contracts

Withdrawal from insurance contracts can be a significant risk for insurers. Withdrawal rates can be difficult to predict because withdrawal is influenced by a number of inter-related factors related to, inter alia, the sales process, characteristics of the insurance contract, characteristics of the contract holder, and economic variables. Existing methods used to model and predict withdrawal rates are initially reviewed. Two additional methods which have been proposed in the literature as means for modelling insurance risks are neural networks and Bayesian networks. These two methods are utilised in order to build models to compare their predictive ability with a commonly used method for modelling withdrawal rates, namely logistic regression.

Grantor dc:publisher.institution

School of Management Studies

Year dc:date.issued

2009

Author dc:creator

• Smith, Bradley

Advisor dc:contributor.advisor

• MacDonald, lain

Language dc:language.iso

eng

Handle dc:identifier.uri

http://hdl.handle.net/11427/5872

OAI identifier oai:identifier

oai:open.uct.ac.za:11427/5872