Teaching an Artificial Central Bank to Conduct Monetary Policy
Type
working paper
Date Issued
2025-04-13
Author(s)
Abstract
This study explores the potential of reinforcement learning to enhance monetary policy design. As an example, I simulate a calibrated, stylised, non-linear economy wherein an artificial policymaker is tasked with setting interest rates to keep inflation close to its goal as well as minimising output gaps and interest rate volatility. The artificial policymaker successfully learns a non-linear decision rule that outperforms benchmark Taylor rules, even under partial information and without knowing the DGP of the economy. The paper illustrates the entire development process, including construction of the simulation, algorithm implementation, policy evaluation and interpretation, to showcase the flexibility and usefulness of the method.
Keywords
E37
E52
E58 Monetary policy
reinforcement learning
Taylor rule
File(s)![Thumbnail Image]()
Name
TeachingAnArtificialCentralBankToConductMonetaryPolicy.pdf
Size
13.67 MB
Format
Adobe PDF
Checksum (MD5)
01413b0a8f7b927fdb89206607426724