Improving cost modelling of residential property replacement costs for short-term insurance purposes: A South African Perspective

Abstract

Listen

Translate article

The residential property market in South Africa has an extraordinarily high number of first-time homeowners. Cost information assistance available to the South African public consists of crude cost models to be found on individual short-term insurers’ websites. The financial cost to obtain an accurate replacement cost estimate from a professional built environment cost advisor outweighs the perceived risk of insuring a residential property for an accurate replacement cost. The need for an alternative cost model that could deliver more accurate replacement costs without employing the onerous cost-estimating techniques as employed in the quantity surveying practice within a short time is apparent. This research aims to develop an alternative approach to building cost modelling for insurance purposes. The building cost model developed, other than that commonly used in the marketplace, is premised on the case-based reasoning (CBR) technique. The four stages of retrieving, reusing, revising and retaining cases are performed. The retrieving incorporates the k-nearest neighbour (kNN) machine learning algorithm to retrieve comparable cost data from a database of residential properties. The database employs the most accurate cost model used in quantity surveying practice and is structured according to recognised building elements. The reusing and revising of the cases are based on specific building features to suit a particular residential property and are performed by applying a mathematical model.The outcome suggests that 75% of predicted replacement costs fall within the acceptable 5% accuracy level of the actual replacement costs, indicating significantly improved replacement cost estimates as the dataset represents costs based on the most accurate cost model used in practice. The study’s findings are important for the South African insurance industry and the built environment as it implies the possibility of providing more accurate insurance values that could curb underinsurance and possible financial setbacks to insureds in future. The findings will also add to the existing generic knowledge on building cost modelling for purposes other than insurance.