Page 60 - India Insurance Report 2023- BIMTECH

48 India Insurance Report - Series II

The premium–income/economic growth relationships are sufficiently strong to be used as the basis
for projections. However, it is worthwhile to take a closer look at dependencies in the error term with
a view to improving the predictive efficiency of the model. Figure (1) given below shows that there is no
presence of auto-correlation as the values of both ACF and PACF did not exceed the UCL & LCL
boundaries. Secondly, LJung Box Q’s values are also quite insignificant, indicating that the goodness of
fit statistics is statistically significant.

Figure 1. Exhibiting the ACF and PACF Indicators of the ARIMA model

ARIMA is a powerful technique with an inbuilt algorithm that weights recent experience differentially
compared to past experience. The application of ARIMA enables us to obtain unbiased parameter
estimates. As we can find from the ARIMA output error statistics, almost all the MSE, MAPE, RMSE,
and BIC values are quite low, suggesting a statistically good fit with the data sets. Results of the ARIMA
output statistics are given in the appendix.