Print this

A handwritten letter from former P&G chairman E.L. Artzt reveals how not to practice risk management

One of the most remarkable documents to emerge from the thousands of pages of litigation between Bankers Trust and Procter & Gamble is a two-page letter written by Edwin Artzt, who served as P&G's CEO when the company lost an estimated $195 million in a series of derivatives transactions.

Included in the undated document is a detailed list of sins Artzt thinks were committed by Bankers Trust and P&G's two senior financial officers: Treasurer Raymond Mains, and CFO Eric Nelson. Artzt says they were notes for a board meeting called to discuss disciplinary proceedings against the two men. "These notes were in my style a way of establishing the credibility of my recommendation by outlining what I saw as the worst case scenario," he explained in a deposition. "In other words, I wanted to say if I were to look at this thing through the darkest glasses to take the severest of actions, these are the things I could say about the situation."

The letter concerns events surrounding a disastrous swap entered into on November 2, 1993, based on a complex formula involving the difference between five-year and 30-year rates. According to court documents, Treasurer Mains signed the contract without reading the formula in it. Mains was informed that the swap was deep underwater sometime in mid-February, and told CFO Nelson soon after. But Artzt and the P&G board weren't told about the disaster until April. Artzt was also not informed about a second swap, based on Deutsche mark interest rates, that was entered into in mid-February, 1994. He blames CFO Nelson for not uncovering the problem until April 5.

Treasurer Mains was asked to take early retirement, but Nelson was retained as CFO, after being censored before the board and denied his yearly bonus. "There was a clear difference between their involvement in this situation," Artzt explained in the deposition. "Nelson did not know about the derivative contracts until after they had been entered into. He was not a participant in the violation of company policy."

IBM Blasts Monte Carlo Models

A new model radically shortens the time needed to perform complex calculations.

On September 27, IBM staffers made an announcement that could very well change the way risk is calculated for years to come. The company has found a way to make cumbersome Monte Carlo simulation thousands of times faster. The breakthrough is based on a new mathematical equation that will be available to the finance community in a new software product known as the Deterministic Simulation Blaster (DSB).

Monte Carlo simulation is a widely used technique for calculating the theoretical prices of complex derivatives and mortgage-backed securities. It subjects a particular financial instrument to a variety of market factors and simulates the price path over a predetermined length of time. Typically, a manager using Monte Carlo analysis will run hundreds - or even thousands - of random price paths to come up with a representative, average path. The more paths that are generated, the more accurate the final, average price - and the more computing power, and time, that is required to run the simulation.

"The Deterministic Simulation Blaster will, all things being equal, provide an accurate price for any instrument between 50 to several thousands times faster than Monte Carlo can," says Ian Colley, program manager at IBM. He explains that this dramatic improvement in speed is due solely to the mathematical breakthrough, and not as a result of ultra-powerful hardware or efficient code.

The system also claims to improve on another associated problem: Monte Carlo models can be more or less accurate depending on the distribution of randomly generated market factors. This is because a "random" sequence does not actually mean that prices are in fact evenly distributed across all the various possibilities. The more random sequences you run, the more likely you are to have a more or less even distribution of prices, but there is always the possibility that your returns will "clump together." Furthermore, it is possible that the same combination of prices appears more than once.

Deterministic simulation, the methodology used by DSB, addresses both the speed and the accuracy problem at once by ensuring that market prices generated by the system are evenly distributed across the possible spectrum and that no duplicate prices are allowed to occur. This even distribution of prices and elimination of duplicates mean that DSB can arrive at accurate price with far fewer "passes" than Monte Carlo analysis. (See chart.)

"Monte Carlo simulation is like a blindfolded man standing on a ledge, attempting to drop basketballs into a grid of shallow manholes," explains Todd Hovanyecz, manager of mathematical analytics and computation at IBM. "Deterministic simulation, however, removes the blindfold and allows our man-on-the-ledge to see where the manholes are, and which ones already contain a basketball." Clearly, the second man is likely to "fill in" his grid much faster than the first, blindfolded subject.

Mark Garman, president of Berkeley-based Financial Engineering Associates, warns users against thinking of the DSB as a silver bullet. "This new form of deterministic simulation may work much better for some instruments than for others," he says. "It is important to test any new model or program thoroughly before moving it into live operations."

Today, IBM is working with a number of its clients to validate DSB's accuracy while running as part of larger, in-house systems using highly individualized pricing models. So far, say the IBMers, so good. Soon, deterministic simulation, IBM-style, may muscle aside Monte Carlo in trading rooms around the world.

End-user surveys: More derivatives, better service

A number of wags have recently predicted that derivatives-related blow-ups would spawn a mass exodus of end-users from the derivatives market. No fewer than three recent studies, however, have concluded exactly the opposite. Although many corporates and investors have reviewed their risk management policies and practices, very few say they intend to eliminate or even curtail their use of derivatives. Consider the following body of evidence:

Based on a study of 168 institutional investors, San Diego-based Intermarket Group Inc. has concluded that funds are not planning to cut back on derivatives usage. In fact, 33 percent of respondents expect to use more derivatives in the future.

After polling over 500 multinational corporate clients, Bank of America discovered that although 70 percent have reviewed their hedging policies and procedures, very few of these internal audits resulted in substantive policy changes.

A Goldman Sachs study of chief investment officers of commercial banks with over $1 trillion in assets and chief credit officers of commercial banks with over $500 billion in assets revealed that the majority of these executives are planning large, net increases in usage for nearly all of the most commonly used derivatives.

So, then, are corporates and institutional investors just bursting with unbridled enthusiasm for their broker-dealers? Not exactly. Many end-users say they are dissatisfied with the level of customer service they are receiving from the dealer community. "Complaints about dealers have been widespread in the past, but I was surprised that there has been almost no improvement in the last two years," says David Strassel, author of the Intermarket study.

Giving Brokers the Shaft

Tough times have done more than reduce spreads in the hypercompetitive interbank swap market - they are inspiring some swap traders to cheat their brokers out of commissions.

Cutting out the middleman is one of the oldest games in business, but here's how it works in the inter-dealer swap market: A swap broker presents a swap trader with a deal from an unnamed counterparty. The trader responds with a price at which he'll do the deal. But when the broker returns with the name of the counterparty, the trader rejects the name - ostensibly because he has reached his credit limit with that particular firm. The trader then calls the counterparty and does the deal directly, cutting the broker out.

"A typical brokerage fee is a quarter of a basis point," explains one insider. "If a dealer can cut out his broker on a $100 million ten-year deal, he can add up to $10,000 to the bottom line of a transaction. If he's getting 30 or 40 percent of a deal, he's putting $3,000 into his pocket."

The scam can work even if the trader never gets the name of the person on the other side of the transaction. "The universe of dealers in longer-term tranactions is finite," adds another. "If they make ten phone calls, they might find the counterparty the broker's showing them."

Brokers have little or no recourse because they are totally dependent on the dealer for revenue. If they threaten to dig up the audio tapes of the conversations, they risk alienating others at the firm. "The best thing to do is simply shut up," says one broker with a sigh.

Who uses RiskMetrics?

The hippest derivatives address on the Internet at the moment is JP Morgan's RiskMetrics volatility and correlation data base (http://www.jpmorgan.com), which has already attracted a loyal following plus a curious fringe group of users.

Although Morgan can't know the exact names of the users with modems who download, their Internet addresses say a lot about who finds the system valuable. In the week of September 25 to October 1, the system logged 1570 downloads: 460 from US commercial users, 187 from educational users, 38 from US government and one from the military. It also listed 77 calls from Germany, 57 from the UK, 50 from Japan and the remainder from 19 other countries.

The biggest users, not suprisingly, were other dealers: 31 calls were from Credit Suisse, 22 calls from Salomon, 18 from Goldman, 15 from UBS, 10 from Lehman and 10 from Morgan Stanley.

Among some of the unexpected were "afterburner.mit.edu" and "blackhole.fnbc.com," along with a couple people at the Vatican. Some of the addresses raise puzzling questions as to who really needs this data. Like why would anybody at the Office of the Supreme Allied Commander of US military forces in the Atlantic (saclant.mil) be a user? Same goes for the Danish defense forces (odin.defence.dnv.no - Odin is the Norse god of war).

Stranger still are the twenty or thirty people who do their downloads not in business hours but on a Saturday morning. Come on nerds! Get a life!