Traditional and Computational Perspectives on Saving Behavior: Evidence from University Students in Azad Jammu and Kashmir”

Abstract

The current study explores determinants of saving behavior of university students in Azar Jammu and Kashmir through the combination of conventional econometric with the latest computational strategy. Four key variables were considered based on survey data and they were examined as financial literacy, parental socialization, peer influence, and self-control. The regression analysis confirmed that all of four variables have a significant and positive impact on saving behavior, the strongest of which are parental socialization and self-control. Machine learning algorithms and  generic random forest were implemented to supplement these results and determine the importance of variables and improve the accuracy of predictions. The most accurate model was the Random Forest model with an accuracy of 84 percent with parental socialization and self-control listed as the most important with peer influence and financial literacy coming second and third respectively. These findings indicate that knowledge, discipline, and the social environment make up the student saving behavior and that computational tools have importance in promoting behavioral finance research. The results have valuable implications to teachers, policy makers, and financial technology creators. To be more precise, they propose the necessity to incorporate the financial literacy programs into the academic curriculum and to design the digital platforms using predictive analytics to help young people adopt improved saving behaviors. This research study  becomes a contribution to the academic discussion and to the solution of concrete problems to encourage responsible financial behavior by integrating the economic theory and machine learning.