Duffie Proposes Stress-Test Matrix to Monitor Risk

The world’s largest banks and investment firms should undergo quarterly stress tests to identify risks that could sink the financial system, according to a proposal by Stanford University finance professor Darrell Duffie.

“I’m not talking about the ordinary, matter-of-course, risk management of institutions,” Duffie said during a presentation today in New York at a conference attended by officials from the Federal Reserve Bank of New York, the Senate Banking Committee, the Commodity Futures Trading Commission and representatives of Deutsche Bank AG and JPMorgan Chase & Co. “We’re looking at what are the sources of risk and how are they flowing through the system. We want to connect the dots.”

U.S. and European governments are working to overhaul the banking system after the housing collapse froze credit markets, helped drive Lehman Brothers Holdings Inc. to bankruptcy and spurred more than $1.8 trillion in losses and asset writedowns worldwide. The U.S. government also had to engineer a $182.5 billion rescue of American International Group Inc. after the insurer sold credit-default swaps to most Wall Street banks in the unregulated private derivatives market.

Duffie calls his plan “10-by-10-by-10” because it’s based on 10 financial firms undergoing 10 stress tests that expose the banks’ 10 largest trading partners. For example, institutions would be tested on their ability to withstand the default of a single firm that they do business with, an idea replicating the 2008 Lehman bankruptcy.

‘Alert Regulators’

“The objective is to alert regulators and the public to potential sources of financial instability before they reach dangerous levels,” Duffie wrote in a paper outlining the proposal.

The tests, which would be adjusted over time to cover different scenarios, could flush out new systemically important firms as they arise, Duffie said.

Central bankers could opt to conduct some of the stress tests using average financial numbers over a given timeframe “to mitigate period-end ‘window dressing,’” Duffie said. Regulators should also audit the way the banks measure the data they present, he said.

Collecting Data

“It’s safe to say we don’t have all of the information,” said Stacy Coleman, a vice president at the Federal Reserve Bank of New York who is leading the central bank’s efforts to curb risks in the privately negotiated derivatives markets. “We have been collecting a lot of information to give us insight into some of these issues, but we don’t have it comprehensively.”

Coleman, who spoke on a panel of regulators and bank and money manager executives to discuss the proposal, said regulators also would need to ensure they have the ability to properly analyze the data.

“Even if we had all the data, I’m not sure we necessarily could run the models in as realistic or likely an outcome as we would want,” she said.

The stress tests should be adopted by all major financial regulators worldwide, Duffie said, adding that the Fed would be a natural choice to lead the examinations in the U.S. because it already monitors the nation’s largest banks. The professor said he sent copies of the proposal to regulators including the European Central Bank, the Bank of England and the Bank of France.

The cost of his plan should be “mitigated by using risk measurement principles that conform whenever possible to best practices in the financial industry,” Duffie said.

JPMorgan Tests

JPMorgan already conducts stress tests that assume rising credit risks, falling stock prices and interest rate changes, Duffie said at the conference, held at New York University’s Stern School of Business and co-sponsored by the Depository Trust & Clearing Corp., which runs a global repository for credit derivatives trades.

The data obtained through stress testing could also be sent to the Office of Financial Research, created through the Dodd- Frank financial reform act that became U.S. law in July, Duffie said. Once analyzed by regulators, the data should be aggregated and disclosed to the public so that financial markets could identify and react to risk concentrations that build up in certain asset classes, he said.

Some of those in attendance expressed concern that identities of bank trading partners could be leaked or otherwise made public.

“The audience seemed a bit concerned if this information should be revealed to the market,” said Viral Acharya, a professor of finance at NYU who moderated the panel.

Anonymous Data

To guard against runs on banks or market speculation that could topple a financial institution, the information released would be “not only anonymous, but in very large buckets,” so that specific firms couldn’t be identified, Duffie said.

Other potential stress tests include analyzing a simultaneous 4 percent change in the yield investors demand to hold all kinds of corporate debt rather than government bonds; a 25 percent shift in the value of the dollar or Euro compared with other currencies; and a 50 percent change in a global stock index.

Duffie said such a monitoring system could encounter problems outlined in the Heisenberg Uncertainty Principle, which states that “increasing the precision of one’s measurement of one aspect of a system merely increases uncertainty regarding other dimensions of the system.”

Thus, employees of a bank may have an incentive to adjust risk-taking to evade standards set by regulators.

“By limiting the stress measures to a small number of extremely broad asset classes, as I have proposed, the ‘Heisenberg uncertainty effect’ is significantly mitigated, but is not eliminated,” Duffie said.

Duffie wrote a research paper in January with Theo Lubke, then the New York Fed official responsible for over-the-counter derivatives oversight; and Ada Li, a New York Fed bank examiner. He is a research associate of the National Bureau of Economic Research’s Systemic Risk Measurement Initiative.

To contact the reporters on this story: Matthew Leising in New York at mleising@bloomberg.net; Shannon D. Harrington in New York at sharrington6@bloomberg.net.

To contact the editor responsible for this story: Alan Goldstein at agoldstein5@bloomberg.net.

Press spacebar to pause and continue. Press esc to stop.

Bloomberg reserves the right to remove comments but is under no obligation to do so, or to explain individual moderation decisions.

Please enable JavaScript to view the comments powered by Disqus.