Principal Investigator:

Dr Mingli Chen

Mingli Chen is Assistant Professor of Economics at the Universiy of Warwick. Her research fields are Econometrics, Time Series Econometrics, Financial Econometrics and Industrial Organization. In particular, her papers are on big data (observations and/or variables are large), detection of structural breaks, large network estimation and the study of consumer payment choice.

Co-Investigators: Andreas Joseph (Bank of England), Michael Kumhof (Bank of England) and Aruhan Shi (University of Warwick).


This project addresses two criticisms faced by the New Keynesian dynamic stochastic general equilibrium model (NK DSGE) since the Global Financial Crisis 2008, namely lack of realistic representation of the financial sector, including its monetary dimension, and the rationality assumption of agents. NK DSGE model is widely used in macroeconomic analysis, and is one of the workhorse models for monetary policy analysis at central banks, including the Bank of England. Addressing the aforementioned criticisms is thus highly policy relevant and applicable. In this project, we address both issues through an NK DSGE model structure with two key building blocks: modelling the financial sector as with money creation and adopting deep/machine learning to model agents’ expectations.

To further illustrate, financial sector, particularly banks, in macroeconomic models are mostly treated as an intermediators of funds between lenders and borrowers. They have no power of ‘creating’ money as people can see in reality, in which banks grant loans for households or firms that are financed through the creation of new deposits, and independently of pre-existing physical savings. In this setting, borrowers’ expectations play a crucial role in affecting their borrowing decisions and financial fragility. Agents’ expectations are more important in affecting how the aggregate economy is affected by disturbances. Naturally, expectation formation is an important building block in this model. Applying deep/machine learning technique, we enable a flexible learning process compared with conventional macro learning models. In this setting, an agent can learn appropriate relation functions given enough training history, which naturally allows for path dependency. Moreover, it has an important technical advantage of being able to solve for a solution in an environment with large action and state spaces. We also address the interpretability of the deep/machine learning methods used.

We want to answer several timely questions through this project. When banks create money in response to shocks, they can do so instantaneously under rational expectations, with potentially large effects for financial stability; how does this change under a different learning approach? What are the solutions of currently used macroeconomic models within a framework of agents that use a form of learning derived from artificial intelligence? Under which conditions is the optimal or rational equilibrium solution obtained. What are the implications for monetary policy, especially in the presence of large shocks which may drive the system to the zero lower bound or into a liquidity trap?

Other than being policy relevant, this project identifies a promising area of future research which has not yet received much attention, namely the intersection between learning in macroeconomics and in the new field of artificial intelligence. As such, the project is also interdisciplinary and aims at the development of new and practical methods.


An introduction blog to this project is available here.


Results will be published here when available.