Can machine learning predict financial crises?
Predicting systemic financial crises has been elusive. It is however a key input for macroprudential policy. We consider the framework of sequential predictions, also called online machine learning, to predict systemic financial crises in a panel of 7 countries. Since there is no consensus on the origins of financial crises, we optimally aggregate a class of reference forecasters using 4 types of aggregation rules. We can predict crisis periods 3 years in advance out-of-sample with remarkable accuracy. We report for each country the subset of variables that appear the most frequently in our best forecasting models.
Answering the Queen: Online Machine Learning and Financial Crises
Authors: Jérémy Fouliard, Michael Howell, Hélène Rey
From: London Business School
Machine learning and liquidity stress
In this paper Heuver and Triepels examine whether supervised machine learning can monitor the payment activities of banks in financial market infrastructures in order to detect early-warning signs of liquidity stress. They construct probabilistic classifiers that estimate the probability that a bank faces liquidity stress. These classifiers are trained on a dataset which consists of various payment features of European banks and spans several known stress events. Their experimental results show that the classifiers detect the periods in which the banks faced liquidity stress reasonably well.
Liquidity Stress Detection in the European Banking Sector
Authors: Richard Heuver, Ron Triepels
From: De Nederlandsche Bank
A review of machine learning methods for economists
In this paper Athey and Imbens discuss the relevance of the recent machine learning literature for economics and econometrics. In particular, the look at:
- The differences in goals, methods, and settings between the machine learning literature and the traditional econometrics and statistics literatures;
- The specific methods from the machine learning literature that could be important for empirical researchers in economics; and
- The newly developed methods at the intersection of machine learning and econometrics. They argue that these methods typically perform better than either off-the-shelf machine learning or more traditional econometric methods.
Machine Learning Methods Economists Should Know About
Authors: Susan Athey, Guido W. Imbens
From: Stanford University
