Predicting Black Swan events using S&P 500

"A Black Swan event is never a question of if it will happen or not? but when it will happen?"

Key properties of a Black Swan event are Unpredictability, Extreme impact and Retrospective Predictability.

Some of the examples of Black Swan Events are

·       Terror attacks like 9/11 and other such incidents

·       Black Monday in 1987

·       Covid 19 in March 2020

·       Global Financial crises of 2008 and Lehman Brothers bankruptcy.

·       And many more ….

Although Black Swan events are unpredictable in nature their impact on the countries, economies and society at large is very high in billions and trillions of dollars, so one must try to identify and predict any upcoming Black Swan event to raise an alert to be prepared for possibility of some unforeseen events.

My research using an approach called Predictive Analytics using Patterns Mining tries to address this problem of predicting Black Swan events using financial markets price data.  The fundamental belief of this approach is that there is a pattern behind each day’s prices of each instrument that is being traded in the financial markets on each day, and if there is a system that can identify the current pattern and checks for the same or similar pattern in the past data, one can know if the future price movement for next one month of a given instrument is expected to rise or fall. Finding patterns is a difficult task equivalent to blind people trying to describe an animal based on touching different body part of an elephant. A pattern has to be derived and checked from different angles.

I instrument I tried to predict is SPX (S&P 500) index to see if SPX is able to predict black swan events from its own past data. There are more than 25000 days of historical price data since this index is being traded for more than 100 years. The findings have been interesting (Partial success and partial failure). I share below the findings.

A)    Current date Oct 15, 1987 (day prior to Black Monday event). The fall in SPX in next one month was -22%. No matching pattern found that could alert in advance.

B)     Current date September 10, 2001 (day prior to 9/11 terror attack event). The fall in SPX in next one month was -10%. A matching pattern was found from May 11, 1970 which had a fall of -10% in next one month. There were multiple reasons for fall in May 1970 like it was last phase of Bear market of 1968 to 1970 and Kent State shootings against students protesting against war in Vietnam and Cambodia.

C)     Current date September 9, 2008 (period of global financial crisis before Lehman Brothers Bankruptcy). The fall in SPX in next one month was -15%. A matching pattern was found from November 16, 1973 when SPX had fallen -11%. There were multiple reasons during the period behind the fall in November 1973 like Arab Israel war, Oil Embargo and Richard Nixon Watergate scandal.

D)    Current date March 4, 2020 (Covid 19). The fall in SPX in next one month was -26% There was no matching pattern found.

Conclusion – Although it will require regular monitoring to see the success ratio, with the limited sample size it had identified and predicted two upcoming major falls. How successfully, concerned authorities are able to make use of it needs to be seen but it could act as an yellow alert to be prepared.

Note –Percentage mentioned are approximate not precise. There are possibilities of adverse event not happening in spite of matching similar pattern. There is also a scope of searching SPX pattern against all other instrument data whose record count is more than 7.5 million records at the time of writing this article. There could be chances of more patterns being found.

About Author _ I am a full time Investor cum Derivatives Trader and a retired IT professional. I have focused on Predictive Analytics using patterns mining after years of research work.