U. S. Subprime Case Study: 2006-2008
Even Nassim Taleb admits that the U. S. subprime crisis was no Black Swan. The macro environment in 2006 was increasingly fragile. Classic macroeconomic imbalances built up over years: low rates and easy credit inflated a U. S. housing bubble amidst record levels of financial leverage. Cracks appeared as the housing market started to taper off in mid-2006, and JPMorgan was the first major bank
Greece CDS Spreads vs 95% Confidence VaR Outliers (100 day moving average)
VaR outliers/100 days
10 Dec 07
5% expected rat* of outliers
Fig. 8 Greece CDS and VaR outliers. Source: Laubsch (2010a, b) to exit subprime. It had already reduce subprime CDO underwriting as credit spreads compressed to leave little margin for error. CEO Jamie Dimon made the final decision to sell all its subprime holdings after observing a spike in subprime delinquencies in its retail bank (Tully 2008).
And yet, complacency still reigned. Credit markets only registered the first small tremor from December 12-21, 2006, when AAA subprime bond volatility tripled. Most market participants did not notice, because the absolute level of volatility was so low. Five year AAA subprime bonds traded around ten basis points (bps) over Treasury’s, and were thus regarded as almost risk free securities.
Daily AAA bond spread volatility around averaged 2 % per day, or about 0.2 basis points. A tripling of volatility only amounted to 0.6 basis points, an increase that was missed by all but the most vigilant institutions which specifically monitored subprime risk. At JPMorgan (where I worked as a risk manager from 1993 to 1998) trading discipline called for meetings whenever VaR limits were breached. At the 95 % daily confidence level, this normally meant such outlier discussions would happen about once month (1 out of 20 trading days). Our objective was to
interpret the market move. Was it signal or noise? When such meetings occurred too often, alarm bells would go off. Risk was not normal and escalating. Goldman Sachs famously decided to “get closer to home” and exit subprime after 10 days of subprime VaR excessions in December 2006. Although absolute losses were relatively low, it quickly became apparent that something was amiss in the subprime world (Nocera 2009).
One crucial insight is that only investors who closely monitored subprime P&L vs. VaR could observe the December 2006 tremors. Firms that mixed subprime bonds with regular bonds would have missed these signals. Hence the importance of defining a Stress Index with specific driving factors.
Figure 9 is a chart of AAA subprime bond yield changes plotted against 95 % confidence VaR bands. Observe that the biggest outlier was February 23, which was more surprising than any outliers observed during the actual crisis.