The topology of card transaction money flows

TL;DR

This paper analyzes the complex topology of real credit card transaction networks, revealing nontrivial distributions and proposing a stochastic model to generate realistic transaction data for better economic modeling.

Contribution

It provides the first detailed analysis of real credit card transaction topologies and introduces a stochastic model based on empirical distributions to improve money flow simulations.

Findings

Real transaction networks have nontrivial topological features.

Transaction distributions follow complex, non-simple patterns.

A stochastic model can generate realistic transaction data sets.

Abstract

Money flow models are essential tools to understand different economical phenomena, like saving propensities and wealth distributions. In spite of their importance, most of them are based on synthetic transaction networks with simple topologies, e.g. random or scale-free ones, as the characterisation of real networks is made difficult by the confidentiality and sensitivity of money transaction data. Here we present an analysis of the topology created by real credit card transactions from one of the biggest world banks, and show how different distributions, e.g. number of transactions per card or amount, have nontrivial characteristics. We further describe a stochastic model to create transactions data sets, feeding from the obtained distributions, which will allow researchers to create more realistic money flow models.