“You’re not going to believe this,” my friend’s note said. “Check out this annual report and look how one of the largest banks in the world is reporting its interest rate risk.” He was right. I took a look at this large and prestigious bank’s interest rate risk reporting and I was immediately transported back to the disco era, 1977. This post turns the clock back 30 years and explains how using net income simulation, with no market valuation-based risk analysis, destroyed the savings and loan industry and cost the U.S. taxpayers $1 trillion the first time. We also explain how a bank, which we call Bank X, traveled so far back in time.
In our August 25, 2009 blog post, we related how the only virtue of being an aging risk manager is that you have a large collection of your own mistakes that you know not to repeat. Not only that, you’ve seen enough others crash and burn because of risk management problems that you remember too. There are two groups of people who don’t know these lessons: (a) younger risk managers, who didn’t have this experience personally and have left behind the principles of best practice finance as “not real world,” and (b) older bankers, senior management and board members who lived these lessons but just forgot them. The phrasing that sent me back to 1977 and my start as an interest rate risk manager at Security Pacific National Bank was this selection from the annual report of one of the largest banks in the world, a bank we’ll call Bank X:
“The table above reflects the pre-tax dollar impact to forecasted core net interest …over the next 12 months from December 31, 2008 and 2007, resulting from a 100 basis point gradual parallel increase, a 100 bp gradual parallel decrease, a 100 bp gradual curve flattening (increase in short-term rates or decrease in long-term rates) and a 100 bp gradual curve steepening (decrease in short-term rates or increase in long-term rates) from the forward market curve.”
That’s all there was, besides the numbers. No discussion of mark to market of assets, liabilities, and “EVE” or economic value of equity. Nothing. So I picked up the 2007 annual report. Then 2006, 2005, 2004, 2003, 2002. The same thing. I tried 1994. Same reporting, and I gave up. How had management forgotten that this approach to interest rate risk caused the U.S. taxpayers to spend $1 trillion bailing out the savings and loan industry? With apologies to those to whom market value based risk management is obviously the right thing to do, we’re dusting off the Bee Gees albums, putting on our leisure suits, and using a presentation not opened since 1993 to explain why this bank is in danger if this is all they are really doing to measure interest rate risk. We also compare Bank X in 2009 to Bank Y in 1996 and explain some of the political history that got Bank X to where it is today: teetering on the brink of serious interest rate risk.
In 1977, I went to work with Paul Smith, the elegant Yale-educated chief financial officer at Security Pacific Corporation. In 1974-1975, Security Pacific had suffered significantly from what with hindsight was a fairly modest spike in interest rates. Paul wanted this problem to be fixed. Security Pacific assembled an asset and liability committee that had some very capable people: Robert T. Parry (later president of the Federal Reserve Bank of San Francisco), Keith Russell (later CEO of Fidelity Savings and Vice Chairman of Mellon), Frank Cahouet (later CEO of Crocker and Mellon), John Kooken (later CFO of Security Pacific), and then-current CEO Richard Flamson. The “resource management department” did the work and it was ultimately led by a very capable senior vice president Gary Updegraft. As I was leaving Security Pacific in 1982, my ultimate co-author on Financial Risk Management in Banking Dr. Dennis Uyemura joined Security Pacific and raised the average IQ by a ton. By then, Dr. John Kooken, a Stanford Ph.D., had become CFO, and savings and loan associations were collapsing on every corner in the wake of a prime rate that hit 21%. The S&Ls were long 30 year fixed rate mortgages and funded with deposits no more than 1, 2 or 3 years in maturity. The smart S&L CEOs knew that their firms were dead and doubled their bets on interest rates on the theory that a small decline in rates might bail them out, and that the FSLIC (now gone, the Federal Savings and Loan Insurance Corporation) would rescue the widows and orphans if the S&Ls failed. The less sophisticated CEOs were less astute than their depositors, who judged that an S&L paying 300 basis points more on its certificates of deposit was a worse deal than the bank across the street, even with federal insurance on the S&L deposit. They pulled out their funds and left it to the FSLIC and FDIC to fire the CEO.
During the 1977-1982 period, the staff at Security Pacific and many other institutions explained to management that a mark to market approach (as popularized in 1973 by the Nobel Prize winning Black-Scholes model and by Frederick Macaulay back in the 1930s—see Macaulay, F. (1938), The Movements of Interest Rates. Bond Yields and Stock Prices in the United States since 1856, New York: National Bureau of Economic Research) was the only risk management technique that worked. By 1984, I had moved to First Interstate Bancorp and the chief financial officer there, Harvard MBA Don M. Griffith, thought the mark to market approach was so obvious that we were given immediate authorization to use that basis for risk management even though the firm’s CEO Joseph Pinola was more of a charismatic leader than a mathematician.
By 1993, mark to market risk management was fully accepted in the industry and a one year (or N year) net income simulation was recognized as absolutely inadequate to protect an institution from complex, long term interest rate risk like that embedded in 30 year prepayable mortgages. Even though that was the case, we still did net income simulations for cash flow and liquidity planning and the CEO’s earnings forecasts for Wall Street. The point here is that it’s necessary but not sufficient for best practice risk management. For example, even the proposed 1993 Federal Reserve interest rate risk regulations used a mark to market approach exclusively. This was thought to be so obvious that there wasn’t even a discussion of net income simulation as a credible alternative. With apologies for the primitive powerpoint look and feel, I dusted off this 1993 presentation (which Powerpoint 2007 couldn’t even open) to explain the state of the art in interest rate risk management at that time and the reasons for it.
Interest rate risk links to capital were first proposed in 1992. A revised set of proposed regulations were announced on September 14, 1993 and published in the Federal Reserve Bulletin in October, 1993.
The presentation went on to explain that the Fed didn’t have the luxury of waiting for the Bank of International Settlements to move on this issue. Congress itself had required new interest rate risk links to capital as part of the Federal Deposit Insurance Corporation Improvement Act of 1991 (FDICIA),. The Fed’s proposals were purely based on a market value approach to risk management, not a net income simulation approach:
The reason for this was obvious. The net income simulation approach, a standard technique, had not worked for the failing S&L industry. This was the early 1980s “oh now I get it” equivalent of the 2009 realization that the copula approach was not effective in dealing with losses in collateralized debt obligations in the 2007-2009 credit crisis.
The Fed’s conclusion was one that few bankers could argue with:
The Fed’s proposed regulations were explained as follows:
Most small banks were excluded from the regulations, unless there was an obvious concentration of interest rate-risky assets:
Banks had a choice of using an “internal models” approach to measuring risk on a market value basis or using a simpler grid-based approach, much like the early Basel capital calculations, to approximate the market value sensitivity to interest rate changes:
Regulators encouraged the use of option-adjusted valuation, and the first version of Kamakura Risk Manager was launched at this time in partnership with Santa Monica based Treasury Services Corporation’s TSER ALM product.
The basis for the Fed’s proposed capital rules was a 2% parallel shift of interest rates:
One of the key issues in making an accurate calculation was the proper treatment of demand deposits and savings deposits (jointly, “non-maturity deposits”):
In fact, this was one of the motivations for Kamakura Managing Director Robert A. Jarrow and me to work on the modern valuation of non-maturity deposits, a topic we published in RISK magazine in 1996. The Fed proposed that banks should be required to add capital if a 2% shift in rates caused market value losses in excess of 1% of assets:
The presentation concludes this way:
Turning back the clock, it’s impossible to overemphasize the consensus among thoughtful risk managers that market value based risk management was essential, because (as shown in another presentation from the same period) banks were failing in large numbers too. It was obvious to all that the industry standard net income simulation techniques were not working to help banks avoid failure.
Back to Bank X: How Could They Not Know This?
Let’s turn back the clock to 2009, leaving the disco era behind. How could Bank X not be reporting on its market value-based risk sensitivity? How could it be relying on disco era risk analysis? Readers of this blog will remember our series on case studies in failures from silo based risk management. We noted in that series that even these failed institutions
- Washington Mutual
- New Century Mortgage
were reporting their market value based interest sensitivity. Our critique was that they stress tested exclusively with interest rates, ignoring other macro factors like home prices, real gross domestic product, and the unemployment rate that were mandated by the Federal Reserve in the February-May 2009 Supervisory Capital Assessment Program. Bank X’s risk reporting doesn’t even meet the standard of these failed institutions. Bank X’s exclusive reporting on 1 year’s net income sensitivity isn’t improved if the maturity is lengthened to 3 years, 5 years or 7 years—one still misses the last 23 years of risk on 30 year mortgages. The contrast between Bank X in 2009 and best practice is made even more stark by this comparison with another large bank, which we’ll call Bank Y.
Bank Y, 1996
To contrast with Bank X’s 1994-2009 concentration on net income simulation as the sole basis for interest rate risk, we select this excerpt from the 1996 annual report of Bank Y. Bank Y at this time was also one of the largest banks in the world. Bank Y makes it clear in this text that it places primary reliance on market value based interest rate risk management, producing net income simulation as secondary calculation that is necessary, but not sufficient, for best practice risk management.
“To minimize exposure to declines in economic value due to gap mismatches, BANK Y's policy is that assets and liabilities must have approximately equal total duration. This asset and liability management policy protects against losses of economic value in the event of major upward and downward yield curve shifts...BANK Y must make a number of estimates and assumptions to calculate its exposure to… risk. For example, BANK Y estimates deposit maturities based on its expectations about future market interest rate movements and its customers' reactions to those movements...
“Overall, management measures … risk by modeling potential quarterly changes in economic value across a range of upward and downward interest rate scenarios that encompass over 99 percent of statistically probable occurrences. This model recognizes that longer-maturity gap mismatches pose greater risk to economic value. Based on this analysis, management believes that BANK Y's maximum economic value exposure attributable to gap positions represented less than 1 percent of common equity at December 31, 1996.
“The near term earnings exposure to gap positions is evaluated in the context of upward and downward interest rate changes assumed to occur ratably over a twelve-month period. This calculation indicated that, as of December 31, 1996, an adverse change in interest rates of 200 basis points would have reduced forecasted annual after-tax earnings by less than 1 percent per year for each of the next two years…
“BANK Y measures options and index risk by developing models based on actualhistorical patterns of changes in interest rate levels, yield curve shapes, and index spreads. Similar to the gap risk analysis, BANK Y must make a number of estimates and assumptions in the process. For example, risk is measured assuming existing embedded options and index mismatch positions will remain open throughout the lives of the related assets and liabilities, a period that can be as long as 30 years. This is consistent with the expectation that loans and deposits will be held to maturity.
“BANK Y then simulates numerous scenarios using a Monte Carlo statistical process. Scenarios generated by this process cover an array of possible future rate levels, yield curve slopes, index spreads, and rates of change in these market characteristics. This process includes a representative number of scenarios that diverge widely from average market expectations. This process ensures that the range and probability of the potential financial effects presented by options and index positions are measured in the context of a comprehensive mix of possible future interest rate scenarios. For each simulated scenario, BANK Y separately models and values cash flows attributable to options and index mismatches. Risk is measured based on the deviation of individual scenario values from the average value of all scenarios.
“Based on this analysis, at year-end 1996, there was a 75 percent probability that BANK Y's loss of economic value would not exceed 2 percent of common equity. At 90 percent and 99 percent probability levels, the analysis indicated maximum losses of 4 percent and 9 percent, respectively, of common equity. Management believes that these adverse outcomes would be distributed over several years and mitigated by intervening risk management actions.”
Epilogue for Bank Y
There have been hundreds of bank acquisitions around the world since 1996, so it doesn’t reveal the identities of Bank X and Bank Y to let you know that sometime after 1996, Bank X bought Bank Y. Given that Bank Y was using state of the art market-value based risk management, shouldn’t Bank X have learned from Bank Y’s risk staff and moved ahead from Bank X’s longstanding reliance on disco-era risk management technology? Yes, that’s what should have happened. What did happen was different. Bank Y’s risk staff and systems were thrown out. Disco lives! The excuse used to oust Bank Y management was ironic. If you want to know what this excuse was, shoot us an e-mail at firstname.lastname@example.org.
What will happen to Bank X? Can Disco survive?
Bank X faces a few scenarios that will bring it out of the disco era in interest rate risk management:
- Traditional pattern, bottom up: Bright risk management staff spends years explaining why net income simulation is necessary but not sufficient. Management finally understands. Market risk management moves to the forefront and net income simulation continues, but with a diminished role
- Rare pattern, top down: Senior management, starting with the CEO and Chief Risk Officer, is replaced. A modern risk staff and modern risk analysis are implemented from the top down.
- Increasingly likely, taxpayers make a change: Regulators step in due to unacceptable risk standards from top to bottom.
Only time will tell which of these will come about. The status quo can’t survive large upward movements in interest rates that persist over time.
Donald R. van Deventer
Honolulu, August 28, 2009