Back to Posts
Listen to Thread

The "August 2007 quant equity crisis" is referenced in http://blog.quantopian.com/meet-quantopians-newest-advisor-wes-mckinney/. Perhaps someone could write or suggest a Quantopian algorithm that would illustrate what happened and why? Specifically, I am wondering if there is a set of predictive indicators that could be derived from the 2007 market data that could be used to flag future crises.

Generally, I'm kinda astounded that more alarm bells weren't going off in the finance community prior to the 2008/2009 meltdown.

Here's one reference that I'm aware of that predicted a potential sharp drop in equity prices:

http://www.efficientfrontier.com/ef/0adhoc/darkside.htm

Hi Grant,

Great question!

Andy Lo (MIT Prof, founder of AlphaSimplex) wrote a paper on the topic of the August 2007 crisis: http://web.mit.edu/alo/www/Papers/august07.pdf Lo hypothesizes that quants had become too highly correlated in their portfolios and risk tools, and that one fund unwinding triggered a snowball effect.

I'm not sure anyone has drawn a connection between the quant crisis in '07 with the credit crash in '08. That said there were some funds that went under in 2007 in an early version of the credit crisis - maybe looking at returns from different fund strategies could be indicative of market regime changes. One similar approach is to track "representative factors", which are almost like trivial investment algorithms. I've heard them called "probe strategies" as well. Based on the simulated returns of these representative factors/probe strategies, some quants try to predict their own algorithms' performance, or the market's behavior.

Example probe strategies: N month mean reversion, N month price momentum, changes in analyst coverage, share buy backs, insider buying/selling, credit rating risk, default risk, small market cap, large market cap, analyst rating revisions, analyst earnings revisions.

thanks,
fawce

Thanks Fawce,

I'd always assumed that there was some connection between the quant and credit crises--perhaps not.

Grant

Hi Grant,

As fawce suggests, it's generally accepted (as of 'now', but new information is always coming to light..) that the two were distinct (although more nuanced interpretations are probably, ultimately, more likely) . What most people recognise as the "quant" breakdown in 2007 was an illustration of what happens with positive feedback in crowded trades when one (big) player has to exit the market rapidly.

(Exit forced by the increased margins calls that counterparties across the market were making at the time. Margin calls were increasing across the market at the time due to losses in the subprime portfolios of those same counterparties - see the nuance? :-) )

The credit crisis was separate to this and triggered completely by lack of short-term funding liquidity (credit) . Literally and quite simply - and I can say this as someone who was standing close by to events - banks would not lend to each other. The short term credit market ceased to function for any size, in any currency, and with any counterparty. Hence institutions had to find alternative was to fund. Liquidating positions was the simplest mechanism, queue fear and crisis.

Thanks Tim,

Sounds like the root cause of both crises may have been the same--too many loans made to too many people who were never going to pay on them.

One thing of interest to me is understanding how so much money could have flowed in obviously the wrong direction without detection. My understanding is that some people saw it coming based on data (e.g. see the link above). So, maybe an algorithm in Quantopian could be devised that would illustrate that a big problem was brewing in 2007/2008 prior to the complete market meltdown? Perhaps it would give some insight into how detect future blow-ups. Any ideas?

By the way, the credit crisis had a pretty immediate impact on manufacturing orders, as I learned first-hand from the planner at my site--at one point, he had zero demand for new products and basically had nothing to plan but service work!

Log in to reply to this thread.
Not a member? Sign up!