Predicting Spending Behavior Using Socio-mobile Features

Human spending behavior is essentially social. This work motivates and grounds the use of mobile phone based social interaction features for classifying spending behavior. Using a data set involving 52 adults (26 couples) living in a community for over a year, we find that social behavior measured via face-to-face interaction, call, and SMS logs, can be used to predict the spending behavior for couples in terms of their propensity to explore diverse businesses, become loyal customers, and overspend. Our results show that mobile phone based social interaction patterns can provide more predictive power on spending behavior than often-used personality based features. Obtaining novel insights on spending behavior using social-computing frameworks can be of vital importance to economists, marketing professionals, and policy makers.