Non-Maturing Deposits: Predictive Modelling and Risk Management

Abstract

Non-maturing deposits (NMDs) are a significant source of liquidity for banks, making research into their modelling and forecasting invaluable. However, NMDs have no explicit expiration date, posing liquidity risks and complicating management. This research develops models and a framework to explain, predict, and manage variations in non-maturing deposits. Aggregate savings and transaction deposit data from an African bank were analysed to test the methodologies. The Trend-Fourier model, leveraging historical trends and Fourier analysis, forecasted 90-day deposit volumes. The model revealed prominent cyclicalities and monthly trends in deposit account volumes. Benchmarking showed high accuracy for savings deposit volumes, while transaction deposit volumes were less accurate, suggesting simpler models might be suitable. Additionally, a risk metric called LVaR (Liquidity Value at Risk) was proposed. Two approaches for calculating the LVaR were tested. An exceedance test demonstrated notable accuracy for savings deposit volumes but struggled with transaction deposits. Results indicated savings deposit volumes were more predictable than transaction deposits. These findings could enhance banks’ balance sheet management by improving non-maturing deposit forecasting. The proposed methodologies could be utilized for internal and regulatory purposes, such as calculating the liquidity coverage ratio under Basel regulations.