Functional Data Analysis for Optimizing Strategies of Cash-Flow Management

Abstract

The cash management deals with problem of automating and managing cash-flow processes. Optimization of the management processes greatly reduces overall cash handling costs. The present analysis is an empirical study of cash flows, from and to bank branches, deriving an underlying theoretical framework, which can in a reasonable way be connected with the optimal strategy. Functional data analysis is considered an appropriate framework to analyze the dynamics of the time series behavior of cash flows: since the observations are not equally spaced in time and their number is different for each series, they are converted into a collection of random curves in a space spanned by finite dimensional functional bases. A central issue in the analysis is describing specific patterns of the curves, taking into account the temporal dependence, and the dependence between curves. The analysis provides a dynamic cash management model that is applied with alternative strategies for programming a cash in transit for the difference between cash inflows and cash outflows in a fixed interval of time. As the strategies are affected by changes in the behavior of the cash flows, the dynamic model outperforms more traditional approaches in identifying the optimal strategy.