Times series analysis of general insurance claims frequency and severity: A case study APA General Insurance Company

Abstract

The purpose of this study was to find out whether time series can be used to predict claims talking APA General Insurance as my case study. Secondary data from published proposal forms APA General insurance were employed throughout the report and analysis at all levels was done at 0.05 significance level using a statistical package known as STATA 13.The randomness test was used for both the number of claims and the amount of claims to test whether each of them are serially independent or occur in a random order. It was done using the runs test. The test for stationarity was done using the Augmented Dickey Fuller test as the test statistic to test whether the amount and number of claims are stationary. The test for seasonality was performed to test whether the amount and the number of claims behave differently over time whereas linear regression was used to test whether the model was a good fit. Claims forecasting was done using Dubin Waston test for white noise/ Autocorrelation for the amount of claims and the number of claims was done using the ARIMA model. From the findings, from the randomness test for the number and amount of claims, a conclusion that the number and amount of claims were random was made. The amount and the number of claims were stationary meaning that there was no trend, seasonality and cyclic effects and that the model used was a good fit. In conclusion, Insurance companies should adopt time series analysis to forecast the amount and the number of claims especially where information about past claims is present so as to know how many claims are likely to be incurred at some point in the future which will hence enable proper planning. I would recommend the government of Uganda to adopt time series analysis through the ministry of finance to engage in business forecasting based on historical trends and patterns so as to boost the economy of Uganda.