Interest Rate Forecasting in the Age of Artificial Intelligence: Economic Projections and Central Bank Practices

Abstract

The interest rate represents the cost of capital, reflects the time value of money, and serves as compensation for the sacrifice of liquidity. As a real and/or monetary variable, the interest rate plays a role in the decision-making of economic agents. In particular, in intertemporal income decisions, it is the ex-ante real interest rate that is used and anticipated, in accordance with Fisher’s fundamental rule. Similarly, central banks forecast their policy rates with the dual objective of stimulating the economy and controlling inflation. The purpose of this paper is to present a theoretical and documentary literature review of the uses of artificial intelligence by these financial institutions, primarily through an analysis of their publications on the subject. Gradient Boosting, LSTM, XGBoost, Random Forest, SVR, Block-Additive, and Boosted Decision Trees are examples of algorithms used by the ECB, the Fed, and the BoE. This sample comprises three major central banks operating in different institutional and macro financial environments, all of which play a significant role in shaping and managing international financial conditions. These AI models have certainly complemented standard statistical and econometric tools, and their ability to refine predictions appears more promising. Nevertheless, they raise ethical concerns related to the risk of increasingly marginalizing the role of humans in monetary policy.