What needs to be said about JP Morgan’s new risk process

As a risk guy myself, JP Morgan has always had a special place in my heart. But the recent news of their reaction to the CDX fiasco has left me somewhat disappointed. Bloomberg news coverage is here:

http://www.bloomberg.com/news/2012-10-12/jpmorgan-deploys-new-var-model-for-cio-bet-reports-lower-risk.html

Now, what’s uncomfortable about this to me is that it entirely misses the point of a.) what went wrong and b.) how to properly mitigate that in the future. Remember from my last post on this, sensitivity to default risk is measured by a metric called CS DV01, which measures how investor perception of default risk (the credit curve) impacts the price of a bond or credit derivative. They had some level of outstanding credit exposure across different parts of their business, and hedged it using a credit index derivative on the IG-9.

Now, there’s a couple important things going on here. As FT Alphaville insightfully pointed out following the meltdown, the CDS-cash basis went haywire around the time of JPM’s loss. Put into other terms, a correlation that is typically stable — the CDS rate on a name and the cash bond credit spread — degenerated. Like all economic variates, correlated credit spreads can at times become uncorrelated. In this case, they did.

And the CDS-cash basis wasn’t the only problem. As I pointed out at the time, what are the odds the IG-9 basket perfectly weighted their duration-adjusted outstanding exposures? Zero. So, they were relying on the correlation of the credit markets as a whole as well. And indeed, this correlation did break down. Again, their fundamental problem was that the stability of correlations is, well, unstable!

Value-at-risk, as a methodology, is built on the idea of correlation. In fact, that’s really all it does — it analyzes the covariance structure between assets in a portfolio to determine the variance of the overall portfolio. So, fundamentally, it is only a methodology that makes sense if you are using it in a situation where correlations can be safely assumed stable. For example, high-frequency timeframes are a good candidate for VaR-style analysis.

But hedging a high-level exposure like this? VaR is simply not a good candidate. Instead, the obvious solution is a detailed and rigorous set of scenario analyses determined by an educated investments team. Target levels should be phrased in terms of these exposures, and limits should be set in terms of the scenario drawdowns. This would allow JPM to set their maximum loss in a manner that is independent of the market’s correlation structure.

I would love to tell you that I’m just a genius that I see all this, but unfortunately that is not the case. I’m quite sure half of JPM echoed this same opinion when asked for their advice. This is more or less Risk 101. Ultimately, Dimon’s reaction seems to be more to the PR impact than the PL impact.

So that leaves us to the final point: put or call on JPM? I believe it is a question of time horizon. An investor looking for short-term gains might be well advised to stick with JPM. After all, public perception influences short-term price behavior more than long-term realities ever will, and Dimon is doing a rather beautiful job of maintaing the appearance of “tough-on-risk,” which may be enough to keep the boat afloat. On the other hand, for the long-term investor, I have to believe this will catch up with them eventually. And the next time that market correlation structures break down, investors may begin to lose faith that JPM is really the safe bet they think it is.

Advertisements

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out / Change )

Twitter picture

You are commenting using your Twitter account. Log Out / Change )

Facebook photo

You are commenting using your Facebook account. Log Out / Change )

Google+ photo

You are commenting using your Google+ account. Log Out / Change )

Connecting to %s

%d bloggers like this: