Is the Claims Ratio Dynamic Predictable? A Study of the Macedonian Non-Life Insurance Sector

Abstract

Insurance, simplified to a risk management concept, offers protection against unexpected losses arising from adverse events. As a financial service of structural importance in modern economies, it requires specific attention due to its risk protection component. Consequently, insurance companies devote sufficient resources towards stable, resilient, and profitable operations. However, a scarce amount of research treats the topic of financial stability maintenance and claims management with even fewer studies dealing with its prediction. This study deals with claims ratio (CR) dynamics in the Macedonian non-life insurance sector and its prediction. Utilizing a data set of 138 monthly observations between January 2012 and June 2023, this paper models the CR indicator through multiple approaches i.e., naïve, ARIMA, ETS exponential smoothing, and random forest (RF) thus making suitability and accuracy comparisons. Results suggest that the SARIMA(4,0,2)(2,1,2,12) model is superior in predicting the claims ratio in both the training and test samples. Moreover, the random forest algorithm shows good performances in the test set but is only superior to the ETS(A,N,A) model.