Developing and comparing machine learning approaches for predicting insurance penetration rates based on each country

Insurance plays an important role in the financial system, impacting various financial variables and vice versa. Many studies have analyzed the long- and short-term relationship between insurance and financial variables, particularly economic growth, using classical and benchmark methods. Given its impact on financial indicators, it is critical for policymakers to predict the insurance penetration rate, which can, in turn, forecast trends in other financial variables. This paper uses decision trees, random forests, and XGBoost to predict the insurance penetration rates for 30 OECD countries, identifying the most effective method for each country. The mean squared error for predicting the total insurance penetration rate with the decision tree and XGBoost is 2.32, and 1.461, respectively, while the loss function decreases to 1.1 with random forest. Additionally, XGBoost outperforms the other models in predicting non-life insurance penetration rate with an RMSE of 0.92, while for life insurance penetration rate, the random forest model has the highest accuracy, with an RMSE of 1.8. Using the prediction, for countries where the insurance penetration rate is going to decrease, policymakers could implement strategies that foster growth in the insurance sector and, consequently, boost other financial variables, particularly economic growth.