Why Risk Models Failed to Spot the Credit Crisis Financial institutions struggling with the subprime mortgage mess all say they conducted "stress tests" to ensure the health of their investment portfolios. But many failed to appreciate the limitations of their risk management models.
NPR logo

Why Risk Models Failed to Spot the Credit Crisis

  • Download
  • <iframe src="https://www.npr.org/player/embed/89507530/89507519" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript
Why Risk Models Failed to Spot the Credit Crisis

Why Risk Models Failed to Spot the Credit Crisis

  • Download
  • <iframe src="https://www.npr.org/player/embed/89507530/89507519" width="100%" height="290" frameborder="0" scrolling="no" title="NPR embedded audio player">
  • Transcript


From NPR News, this is ALL THINGS CONSIDERED. I'm Robert Siegel.


And I'm Michele Norris.

Finance ministers from the world's seven richest nations are meeting in Washington today. They're expected to announce a new team of supervisors to watch over the world's biggest banks. And there are a lot of people saying regulators need to keep a closer eye on things. The global credit system is a mess after a series of major U.S. and foreign banks made bad bets on subprime loans.

NPR's Adam Davidson looks at how those firms could have gotten things so wrong.

ADAM DAVIDSON: Every big bank has a risk management team whose job it is to keep the bank out of trouble. They use complex computer systems to guide them away from financial danger. But as we've seen, those systems didn't work very well for a lot of banks.

Andrew Aziz works for Algorithmics, which makes some of the more popular risk management software. He told me to look at an online demo.

Dr. ANDREW AZIZ (Executive Vice President of Risk Solutions, Algorithmics): So what you should see are a collection of reports.

DAVIDSON: Wait, hang on one second. I just clicked on this link in the e-mail you sent me and I typed in the secret code that you gave me. Here, up pops AlgoRisk 2.3. It looks like Quicken - there are graphs and pie charts and spreadsheets, but it functions more like a hysterical new parent obsessively running through countless scary scenarios. What if there's a recession? What if interest rates fall too far or rise too fast or the central bank of Kazakhstan collapses?

It studies hundreds of thousands of these scenarios every day, trying to figure out how likely a bank is to lose a lot of money. And for all its worrying, the software completely failed to predict the current subprime crisis. Aziz says that to understand why the software failed, you have to understand the basic mathematics of risk management. And to do that, you have to go somewhere unexpected.

Dr. SUSAN PELL (Molecular Plant Systematist, Brooklyn Botanical Garden): Would you like to collect the pollen?

DAVIDSON: It turns out that to understand why banks were so bad at predicting this crisis, it helps to visit a botanic garden. In this case, the Brooklyn Botanic Garden and Dr. Susan Pell.

Dr. PELL: Here you go, here are some twitters. You can just climb over here and maybe get down on that rock bed down there. Here, I'll hold your hand.

DAVIDSON: Financial risk management uses a model discovered in 1827 by the grand Scottish botanist Robert Brown, when he did exactly what we're doing. He collected some pollen and then went back to his lab and put the pollen and some water under a microscope. He saw, just like we saw, that the tiny pollen grains bounced around completely randomly, buffeted by constantly moving water molecules.

Dr. PELL: So this one right now is sort of jumping up and down, and it's gone to the left here and now it going up to the right again and, sort of, the whole time jiggling back and forth as well.

DAVIDSON: It turns out that what Brown discovered, we now call it Brownian motion, is useful in studying all sorts of random things including financial risk. It's useful, not foolproof.

Richard Lindsey wrote "How I Became a Quant" about financial engineering. He says financial models built on Brownian motion have done a lot of good. They are so helpful in managing risk, at least some of the time, he says, that they have made the world far richer. But that doesn't mean they always work.

Mr. RICHARD LINDSEY (Author, "How I Became a Quant: Insights from 25 of Wall Street's Elite"): I think that it's fair to say that it's not perfect in modeling in normal times.

DAVIDSON: Why do we use it if it's not perfect?

Mr. LINDSEY: Because it's about the best thing that we have.

DAVIDSON: Brownian motion works best with truly random things, little bits of pollen that bounce around with no rhyme or reason. Financial markets are not random; they are linked. But that financial software is not always good at figuring out how those global links work.

Andrew Aziz, who helped write the software, said they know all about these limitations. He says the software is a great tool, but it can't do everything. He says it's like a GPS device in your car.

Dr. AZIZ: A GPS system helps you get from A to B in a car in a much better manner, but no one would rely strictly on the GPS and not look out the window.

DAVIDSON: I wonder if AlgoRisk shouldn't have a warning: Risk may be larger than they appear.

(Soundbite of laughter)

Dr. AZIZ: We just might do that.

DAVIDSON: Aziz says some banks know all about Brownian motion and the vulnerabilities of the software. They planned correctly and are doing well now. Other banks just never looked out the window and they crashed.

Adam Davidson, NPR News, New York.

Copyright © 2008 NPR. All rights reserved. Visit our website terms of use and permissions pages at www.npr.org for further information.

NPR transcripts are created on a rush deadline by Verb8tm, Inc., an NPR contractor, and produced using a proprietary transcription process developed with NPR. This text may not be in its final form and may be updated or revised in the future. Accuracy and availability may vary. The authoritative record of NPR’s programming is the audio record.