JPMorgan Silence on Risk Model Spurs Calls for Disclosure
JPMorgan Chase & Co. (JPM)’s multibillion- dollar trading loss exposed an industry practice that U.S. regulators are now likely to clamp down on: Banks keep investors in the dark about how they calculate trading risks.
The U.S. Securities and Exchange Commission is probing JPMorgan’s belated May 10 disclosure that a change to its mathematical model for gauging trading risk helped fuel the loss in its chief investment office. While the SEC would have to prove that the biggest U.S. bank improperly kept important information from investors, regulators probably will press Wall Street firms to tell more about the risks they’re taking, three former SEC lawyers said.
“The SEC and investors are learning that apparently the compliance folks and financial folks at very senior levels at JPMorgan do not think it’s important to share really significant modeling changes,” said Elizabeth Nowicki, a former attorney in the SEC general counsel’s office who’s now an associate law professor at Tulane University in New Orleans.
So far, New York-based JPMorgan has added a warning in its most recent quarterly report that risk models are continually tweaked to account for “improvements” in modeling techniques, and the head of the SEC has publicly asserted that banks should disclose significant changes and the reasons.
The dispute revolves around value-at-risk, the main and sometimes only empirical gauge that investors get as they try to fathom how much a bank could lose if its trading bets go bad. Wall Street firms routinely give only broad outlines of how their mathematicians calculate VaR, according to data compiled by Bloomberg, and almost nothing about changes in statistical assumptions or the prices they choose to feed into their models.
The skewed comparisons can leave investors guessing about whether the potential for loss is rising or falling, according to risk analysts. Adding to the muddle, some bankers including JPMorgan Chief Executive Officer Jamie Dimon have told investors not to rely too much on VaR, calling it just one part of their effort to manage risks at the biggest lenders.
SEC Chairman Mary Schapiro told a Congressional panel last month that her agency is scrutinizing the banks’ disclosures about the change in VaR models for its chief investment office, which invests the banks’ excess cash. The Federal Reserve and Office of the Comptroller of the Currency are examining why the change was made and whether it’s a sign of weakness in risk- management practices.
They’re asking because JPMorgan changed its calculation methods on Jan. 15 for the CIO as the unit was grappling with how to unwind stakes held by Bruno Iksil. The U.K.-based executive was dubbed “the London Whale” after his trades became so big that they distorted prices in an illiquid market, making it hard to get out of the positions.
Losses on the trades may have climbed to $4 billion in the second quarter, according to John McDonald, an analyst with Sanford C. Bernstein & Co., far exceeding the amounts predicted by JPMorgan’s VaR gauge.
Value-at-risk represents the maximum trading loss that might be expected by JPMorgan on 95 out of 100 trading days. The SEC ordered banks to disclose information about their market risks through a rule adopted in 1997.
JPMorgan didn’t tell investors about the change in its formula before it took effect, or in its first-quarter earnings report on April 13, when it said the figure averaged $67 million. It wasn’t until May 10, after losses mounted to about $2 billion, that the bank disclosed that the formula had changed -- and that it had been discarded in favor of the old one, which showed average VaR was actually $129 million. JPMorgan’s shares have tumbled 17 percent since then through yesterday, the worst showing among the four biggest U.S. banks.
Dimon, 56, told a Congressional panel last month that his bank has hundreds of computer-based models for estimating trading risk and they’re constantly updated. He said that the change in the VaR model and its lower reading “did effectively increase the amount of risk that this unit was able to take,” adding he didn’t believe it was altered for “nefarious purposes.”
There’s “tremendous subjectivity” in how VaR is calculated, said Aaron Brown, a former Citigroup Inc. (C) and Morgan Stanley (MS) executive who’s now head of risk management at hedge fund AQR Capital Management LLC in Greenwich, Connecticut. In his 32-year career, Brown said, he had never seen a bank disclose a change in the risk model for a specific trading desk.
“What would you tell people, that we changed this parameter that we never told you about in the first place?” Brown said. The threshold might depend on the potential impact, he said. “Maybe it was so big it should have been disclosed.”
Not once in the three years before JPMorgan’s trading loss did the bank disclose a change in any risk models for specific trading desks, according to data compiled by Bloomberg.
Underscoring the point, JPMorgan’s quarterly report on May 10 included a disclaimer -- for the first time -- that risk models are changed frequently.
“VaR models are continuously evaluated and enhanced in response to changes in the composition of the firm’s portfolios, changes in market conditions and dynamics, improvements in the firm’s modeling techniques, system capabilities and other factors,” the bank said on page 73 of the 178-page report.
Jennifer Zuccarelli, a JPMorgan spokeswoman, declined to comment on VaR, citing the bank’s plan to discuss the CIO’s losses in conjunction with a July 13 second-quarter earnings announcement. John Nester, an SEC spokesman, declined to comment on the agency’s investigation.
JPMorgan isn’t alone in tweaking its risk models behind the scenes or employing discretion to calculate the figure. Goldman Sachs Group Inc. (GS) and Citigroup also make changes to their VaR models regularly, people with knowledge of the matter said.
Schapiro, 57, testified before the House Financial Services Committee on June 19 that banks’ quarterly reports are supposed to include “discussion and analysis” of any “material changes in the market risk.”
They must also “disclose any changes to key model characteristics and to the assumptions and parameters used, as well as the reasons for the change,” she said. “Changes to the scope of the instruments included within the model and the reasons for those changes must be disclosed as well.”
Value-at-risk is critical enough to investors that analysts including Guy Moszkowski, formerly of Bank of America Corp. (BAC), and Glenn Schorr of Nomura Holdings Inc. asked Dimon during at least three quarterly conference calls about changes in the figure.
Dimon, responding in April 2009, said that he didn’t “pay that much attention to VaR.” In January 2010, he told analysts that it was “really not an accurate measure of risk.” JPMorgan uses data going back one year, Dimon said, so the figure might fall just because a volatile week was no longer included in the calculation. He didn’t mention then that the models themselves were constantly changing.
“It was exceedingly difficult for third-party analysts to diagnose the magnitude of JPMorgan’s CIO hedging portfolio risk buildup based on bank-provided disclosures,” David Hendler, an analyst at CreditSights Inc., wrote in a June 18 report. Only “on-the-ground” hedge funds trading credit derivatives could discern the buildup, he wrote.
JPMorgan supplements its VaR modeling with other risk- management tools, including “stress tests,” in which bank officials estimate how trading positions might perform in deteriorating economic conditions or market turmoil, according to the bank’s annual report. The results of the internal stress tests aren’t disclosed.
Harvey Pitt, a former SEC chairman, said the agency might allege that when JPMorgan adopted the new VaR model in January, it should have disclosed any perceived defects with the old one -- and how results might have differed with the new model.
“The SEC would be concerned about how JPMorgan became aware of the original problems, how it satisfied itself that it had repaired those deficiencies and how much assurance it had that its new model would perform,” Pitt wrote in an e-mail. “JPMorgan did not bother to address these issues.”
Based on past regulatory practice, bringing a case against the bank may not be warranted, said Edward Greene, a former SEC general counsel who later served as legal chief for Citigroup’s trading and investment-banking division.
“All these banks have a VaR approach, and they come up with a range of values that they disclose, but to my knowledge the SEC has never required you to give much background as to how you change it or adjust it,” said Greene, now senior counsel at the law firm Cleary Gottlieb Steen & Hamilton LLP in New York. The SEC “may have to show that there would be some intent to deceive the market, or reckless disregard.”
Banks typically disclose only changes to the broadest parameters of their risk models. In 2008, for example, JPMorgan switched its VaR formula to use the 95 percent “confidence level” from 99 percent, according to the bank’s annual report for that year. The move, which reduced the bank’s year-end VaR to $286 million from $317 million, was intended to “provide a more stable measure,” the bank said.
Goldman Sachs changed its VaR model in 2004 to exclude “distressed asset portfolios” that couldn’t “be properly measured,” according to the bank’s annual report filed in February 2005. The impact of the change was “not material to prior periods,” so previously reported figures weren’t adjusted, New York-based Goldman Sachs said. The company’s recent filings haven’t disclosed changes to any VaR model for a particular trading desk.
Morgan Stanley, owner of world’s largest brokerage, has included a disclaimer in its annual report since at least 1998 stating that its VaR model “evolves over time” in response to “improvements in modeling techniques and systems capabilities.” Bank of America’s annual report for this year included the line, “We continually review, evaluate and enhance our VaR model.”
Citigroup said in its annual report its VaR model is updated weekly. In the first quarter, a 10 percent drop in the New York-based bank’s average VaR compared with the prior three- month period was “primarily driven by a change in VaR model parameters,” according to a May 4 filing.
Jon Diat, a Citigroup spokesman, said the firm “provides extensive disclosure on value-at-risk.”
Risk modelers and mathematicians at banks routinely make choices that affect the VaR calculation, said AQR’s Brown, who in February was named “risk manager of the year” by the Global Association of Risk Professionals, a trade group. For instance, banks can base their risk estimates on absolute numbers or percentage changes.
“That is something that might change without somebody noting it in an annual report,” said Steve Allen, a former head of risk methodology for JPMorgan who retired in 2004. “It could have made a really big difference.”
When calculating VaR for an asset without a complete pricing history, traders and risk managers have to choose a “proxy” -- an asset whose price moves are believed to be similar to those of the asset being modeled.
“The choice of proxy is a subjective judgment,” said Allen, now a consultant based in New York.
Goldman Sachs regularly tweaks its models, Chief Financial Officer David Viniar said in May at a Citigroup-sponsored money- management conference in Napa County, California, according two people with knowledge of the remarks. What’s unusual is for a model change to result in a doubling of the VaR, Viniar said, according to the people, who asked not to be identified because the session wasn’t public.
Viniar added that JPMorgan executives are good risk managers, one of the people said.
The SEC “could allege that the books and records were not accurate, or that they failed to disclose material facts or amend timely obligatory disclosures,” said Jacob Frenkel, a former SEC enforcement lawyer who’s now at the law firm Shulman Rogers Gandal Pordy & Ecker PA in Potomac, Maryland. “But it is a terrible precedent if the SEC starts second-guessing or expecting that every tweak of a model or a calculation for legitimate reasons should somehow implicate a violation.”
To contact the reporter on this story: Bradley Keoun in New York at email@example.com
To contact the editor responsible for this story: David Scheer at firstname.lastname@example.org