Since I’ve become something of a volatility junky, let me point you to this interesting FT article that describes how volatility is wreaking havoc on the way financial firms value their assets. An excerpt:
The second, more tangible implication of the return of volatility relates to the models that banks and hedge funds use to measure their assets. When banks extend credit to hedge funds, they often use so-called “value at risk” models (VAR) to measure the risks attached to such loans. These models typically assess the riskiness of an asset by measuring how its market price has moved in the past.
During the Great Moderation, this approach cast a fabulously flattering light on the investment world, creating the impression that it was safe for banks to extend massive volumes of credit to hedge funds. Moreover, since banks typically use VAR to measure the risk attached to their own assets too, these models also seduced banks into feeling complacent about their own risks.
Now, this process has gone violently into reverse: as volatility surges, VAR models are showing that the risk attached to almost any transaction has exploded upwards. Thus banks are selling assets and slashing loans to the funds – in turn sparking more fire sales and increasing volatility in all asset classes. It is a vicious trap that does much to explain why the market upheavals have infected so many asset classes, ranging from subprime to sterling to Shanghai shares.
I’ve been feverishly working on a magazine piece about risk, and yesterday I happened to have a conversation about the VAR models that were used to value mortgage-related structured products. The guys I was talking to kept saying that a big problem was that VAR models used during an asset boom make the future seem extra-rosy since the model extrapolates from that historical data. I asked why firms didn’t use more historical data, maybe going back 30 years instead of–what?–ten years?
Um, no. Try two or three. So we were deriving values for things like CDOs from just a few years of historical data. It’s not that the VAR models captured the boom—it’s that they only captured the boom. Brilliant. And then plenty of firms either weren’t stress testing those models, or they were, and deciding to ignore the results. But more about that in a week or so.
Barbara!