A discrete choice approach to model credit card fraud

This paper analyzes the demographic, socio-economic and banking-specific determinants that influence the risk of fraud in a portfolio of credit cards. The study is based on a large banking portfolio and is the first to analyze the statistical significance of the determinants of credit card fraud in Italy. A logit framework is employed that incorporates cards at a risk of fraud as the dependent variable and a set of control variables (e.g., gender, location, credit line, number of transactions in euros and in non-euros currency). The empirical results provide useful indicators on the factors that are responsible for potential risk of fraud. Over all, the riskier categories are women, customers residing in the Center of Italy, those who own a II circuit credit card and secondary owners.