Zhuohui Chen, Nikolaos Kourentzes, and Mr. Romain M Veyrune
Implementing monetary policy largely consists in controlling short-term interest rates which supposes having a good understanding of banks’ demand for liquidity also called “reserves” at the central bank. This work aims to offer a modeling methodology for estimating the demand for reserves that itself is influenced by various macro and market structure variables. The model can help central banks to identify ”stable points” on the demand for reserves, which correspond to the levels of reserves for which the short-term interest rate volatility is minimal. Both parametric and non-parametric approaches are provided, with a particular focus on capturing the modeling uncertainty and, therefore, facilitating scenario analysis. A method is proposed to test the forecasting performances of different approaches and exogenous regressors combination, finding that simpler parametric expressions provide on balance better performances. Adding variables to both parametric and non-parametric provides better explanations and predictions. The proposed methodology is evaluated using data from the Euro system and the US Federal Reserve System.