Julia Holmgren: Segmentering av reservens påverkan på mean squared error of prediction

Abstract.

The purpose of this study is to evaluate whether the precision of reserving is improved by setting insurance reserves for an entire portfolio or by calculating the reserves on a risk level. This is measured by comparing the Mean Squared Error of Prediction (MSEP) on simulated data for the reserves. The data is simulated based on real home insurance data from an insurance company and has a similar mean and standard deviation. The reserves are calculated using the chain ladder and development factor methods. The results show a slightly better result when using the risk level segmentation compared to the entire portfolio. This implies that using a segmentation when calculating reserves could be preferrable when measuring MSEP. Further studies could for example validate these results by examining and comparing with additional segmentations or evaluate using truncation to reduce noise.