3 New Ways to Predict Financial Crises

Albert Edwards, an economist at Société Générale, a banking and financial services firm, forecasts a new financial crisis by the end of 2016. In his opinion, it will be as bad as the one in 2008–2009. “Sell everything except high-quality bonds. This is about the return of capital, not return on capital,” Andrew Roberts, Head of Credit at RBS bank, advised investors gathered at an investment conference in London last January. Investor George Soros believes that the main cause of the upcoming banking distress will be China’s economic decline. It will be “[…] exporting deflation to the rest of the world”, he explained at the World Economic Forum (WEF) 2016, held in Davos, Switzerland, in January.

Such public statements of 3 renowned figures in finance is a warning that the global financial system is at risk. However, there are other factors to consider before raising the alarm. There are many market players who affect each other. Very often their business interests collide. Under certain economic circumstances, such collisions can disbalance the entire financial system. In order to prevent this, new forecasting methods are being developed.

Deep learning algorithms

Computer scientists Samuel Rönnqvist and Peter Sarlin have created a deep learning algorithm which recognizes data patterns in financial news indicating upcoming financial difficulties. Having processed 6.6 million financial news articles from 2007 to 2014, it precisely identified moments of banking distress, and found a text extract which explained the most likely causal relations between them. The scientists believe that such algorithms will soon be used to analyze real-time data (RTD) to forecast financial crises. Special attention will be paid to the emotional aspects of public speeches.

Sarlin and his colleague Markus Holopainen have already analyzed how accurately algorithms can spot conditions which lead to crises. They combined different machine learning methods to process macroeconomic data of 15 European countries since the 1980s. They found out that crises were predicted with better accuracy than when using regular statistical methods.

Agent-based model No. 1

In Science Magazine, a group of financial experts has proposed developing a model which would better reflect multiple viewpoints and relationships in a financial system. It would be based on complexity theory, which claims that there is a pattern of how complex systems are likely to evolve. “It’s about the notion that the interaction of simple parts can give rise to behaviors that are qualitatively different than the parts themselves,” explains J. Doyne Farmer, an External Professor at the Santa Fe Institute.

The new model would overcome a common mistake in traditional models: it would not rely on an average economic agent but many individual agents. Developers would assign some characteristics to each of them and some value to all their possible decisions. An algorithm would suggest which decisions they are more likely to make in a certain situation. The agent-based model (ABM) would be especially useful for testing new policies before implementing them.

Agent-based model No. 2

A group of researchers at the US Treasury Department’s Office of Financial Research (OFR) led by Richard Bookstaber is developing an agent-based model which analyzes which actions of individual agents can lead to financial crises. Each of them is assigned a ‘decision rule’ based on their priorities, financial position, behaviour patterns, relationship with other agents, actions they might take in a given situation and their implications. The model simulates real life situations to see how the interaction between the key agents will evolve. The process is repeated under different scenarios to get a complete picture of ‘crisis dynamics’, as Bookstaber calls it.

Once the team of experts starts working with live data, it will be able to make the model better reflect real life interaction. Although the project will be completed within a few years, agent-based models are already being used to examine the effect of large-asset liquidation on the market and the stability of financial networks, reveals OFR spokesman William Ruberry.

References: [1] Joseph Conroy, newstalk.com / [2] Stuart Burns, MetalMiner / [3] Mark Buchanan, BloombergView / [4] Amy Nordrum, International Business Times / [5] Leah McGrath Goodman, Newsweek