Exploring customer Spending Behavior and Payday Effect using Prepaid Cards Transaction Data

Background: Prepaid cards are reloadable charge cards typically designed for customers that would not be able to have a traditional credit card due to no or poor credit history, and hence unable to open checking account or credit card at a bank. Many prepaid card customers would have their salary credited to their account and make most withdrawals and purchases on this card since they do not have many other options. Their data are therefore very suitable for studying customer spending behaviors. Aim: In this work, we study how income affects customer spending patterns. Specifically, we aim to determine whether or not a “payday effect”, i.e. payroll incomes affecting customers’ spending behaviors, exists in the prepaid card customers and if so how does it vary over time and transaction categories. Data: We use prepaid card transaction data collected by a large national bank over the course of 28 months which covers customers across the United States. The subset we considered contains 41,610 customers. We processed the data so that there are 6 categories of spending and 4 categories of deposits. The length of the transaction history of each customer varies between 11 days to 28 months. Method: We use a variant of the linear regression model from prior studies on payday effects and carefully control for possible confounding factors, specifically balance features. Results: Strong (and statistically significant) responses are detected for all transaction categories on the coefficients corresponding to indicator features for days on and after payday. The explained variances on non-essential spending (restaurants and others) are larger than that on essential spending (grocery, gas, etc...). Conclusion: We conducted our analysis on the new prepaid card transaction data with additional control for balance. A strong payday effect exists on prepaid card customers despite the predictable nature of payroll income. The strengths may vary across different types of spending.