Mental Accounting and Savings Behavior: Evidence From Machine Learning Method

Abstract

This article proposes a machine-learning-based method that can predict individuals’ savings behavior in the presence of mental accounting. The proposed predictive model perceives wealth and consumption as each being divided into three nonfungible distinct classes, as posited by the mental accounting theory. The predictive model found that mental accounting categories do have predictive power on savings behavior where expenditure on discretionary items and current income are most important. Savings behavior is best predicted by the random forest model based on the Malaysia Ageing and Retirement Survey data.