Description:
This paper proposes a unique approach to simulate intraday transactions in the Canadian retail payments batch system. Such transactions are currently unobtainable. The simulation procedure, though demonstrated in the realm of payments systems, has tremendous potential for helping with data-deficient problems where only high-level aggregate information is available. The approach uses the concept of integer composition in combinatorics to break down the daily total value and volume (available) into individual data points (unavailable) throughout the day. The algorithm also introduces a technique to incorporate any intraday timing pattern (known or hypothetical) to make simulated data closer to reality. Simulation results show that the probability distribution of individual payment values is remarkably stable through repeated random sampling. This suggests a high degree of accuracy and viability of this simulation method. In addition, the densities of simulated intraday transactions are found to be invariably skewed to the left of the mean payment value, which reflects the nature of retail payments systems.