Let us agree that the risk of being in a movie theater in a town where some nut has been crying “Fire” in movie theaters can be expressed as:
R=f(ND, WD, WA, NP, AP, DFE, NPBYA)
Where ND is the Number of Doors;
WD is the Width of the Doorways;
WA is the Width of the Aisles;
NP is the Number of People in the Audience;
AP is the Athleticism of the People in the Audience;
DFE is the Distance From the Exit of the seat you occupy; and
NPBYA is the Number of People Between You and the Aisle.
The only factors under your control are DFE and NPBYA, and all of the factors are S-curves, with a steeply ascending measurement of danger beyond a threshold. You cannot determine how urgently you should consider ND, WD and WA, or how much weight you should put on DFE or NPBYA, unless you know NP. If the theater is so dark that you cannot tell the size of the audience, you live in a world of uncertainty, where the calculation of probabilities by mathematical means is analogous to the drunk looking for his lost keys under the lamppost.
In the theater of over-the-counter derivatives contracts entered into behind closed doors, with no register of open interest, you cannot know NP, you cannot know whether ND or WD have kicked into the equation, and you cannot even guess the probability that when you go to the movies you will get out alive. A door that would be wide enough for you to pass through even if you are the tenth of 10 people rushing for the exit would be a killer if a hundred people were trying to escape at once. And in the derivatives context, there is a further problem that the more popular the analysis, the bigger the crowd—and the more likely that leveraging will fill the aisles with fat people at the first whiff of smoke. By the time the Fed bulldozes the wall to get everybody out, there are going to be lots of dead.
As an alternative analogy, we might consider Andrew Sheng?s description of “financial markets as networks,” in his address at the Conference on the New International Financial Architecture in Hong Kong last June. “Volatile capital flows can be thought of as power surges across the network,” he said. “Those hubs in the global network that were designed to take only four standard deviations in volatility when a 15 standard deviation volatility occurred would blow a fuse.”
The period 1997-99 has been remarkable for its events, the verbal reaction to those events, and the inability of the theater managements and the fire departments to do anything about the dangers the events exposed. The G22 Working Group on Transparency and Accountability stressed the need to close gaps in information disclosure—and also the potentially more important gaps in the incentives to use the information disclosed. Transparency is the watchword. We must turn up the lights in the theater so that all who enter have a good idea of the NP, and can decide where they want to sit, or even whether the movie is good enough to justify the dangers.
But in fact the central bankers continue to turn down the lights. They learned nothing from the Hunt fiasco, now 20 years old. More than a year after the collapse of the Russian markets, we still do not know the volume or cost of the non-deliverable forward contracts that banks, dealers and end users wrote for each other awarding dollars to hedgers who feared the decline of the ruble. In the fall of 1999, we don?t know who that ain?t got gold is short gold, and how much And—blessings upon us—we have unknown quantities of total-return swaps, where all the pleasures and pains of equity investment can be experienced by people with enough credit to borrow U.S. Treasury paper.
The Bank for International Settlements still promotes the bilateral settlement of derivatives, which takes trading in derivatives out of the clearinghouses and into correspondent banking relationships, notoriously the most opaque arrangements in finance. And the Federal Reserve still takes the position that convergence trading improves price discovery, and resists the imposition of risk weightings that would make such activity (and its financing) unprofitable. Of course, the price outsiders discover is at best the ISDAFIX, which takes the rates reported to Reuters by 16 banks, throws out the top four and the bottom four and averages the middle eight.
The favors done to convergence trading are perhaps the most remarkable aspect of the failure of regulatory authorities to come to grips with the realities of the markets. Convergence trading is simply one way to sell volatility. Like the gambler who goes to the roulette wheel and doubles his bet until he wins, the convergence trader has a sure thing—provided his financing continues at a price he can afford when the rates or values that are supposed to converge separate still farther. As Merton Miller has pointed out, if Deutsche Bank hadn?t been such wimps about a few billion dollars more until the usual backwardation resumed, Metallgeselschaft would have made money.
Because popular preferences are positively skewed—because most people prefer a large probability of losing a small sum and a small probability of winning a large one to a situation where you are most likely to win small but can lose big—the pricing of volatility tends to be off center. And in the construction of systems like RiskMetrics, the analysts weight price changes more heavily according to their recency, which means that violent upheavals drop out of the database relatively quickly. Again, because the market anticipates what the machine has been programmed to forget, the pricing of volatility is off center.
Thus the seller of volatility is highly likely to show small profits nearly every day. And because of the air of “certainty” in the process, he can leverage his positions dramatically. The President?s Working Group on Financial Markets has estimated the leverage of hedge funds at 25 to 1, and Long-Term Capital Management, the fund most committed (in theory) to convergence trading, was leveraged even more highly. But, as George Soros has pointed out, “The value of collateral is influenced by the availability of credit.” And Andrew Sheng takes the analysis a step further: “Assets financed mostly by leverage have higher volatility of values,” he said in June. Thus the anecdotal observation of Chicago traders—that sellers of volatility “eat like chickens and shit like elephants”—in fact expresses a practical certainty with theoretical justification.
When John Meriwether of LTCM looked back on his experiences, he told potential funders of a second go-round that his problem had been that he had no notion how many other people were following the same strategies. But imitation is not only the sincerest form of flattery—it is also the most likely form of flattery. Having touted the movie highly, the volatility seller cannot be really surprised that NP is high, there are lots of people in the theater. But because both he and his imitators have operated mostly in bilateral opaque markets—and have indeed sought opacity—none of them can know how crowded the theater is. Nassim Taleb tells a story of a trader who tried to protect a profit by selling volatility just before an explosion in the market, and who “concluded with the following wisdom: Hedging increases your risks.”
It should be noted that while regulators held their fire—no actions were taken on the pioneering recommendations of the committee on highly leveraged institutions of the Basle Committee on Banking Supervision—market participants began to require collateral, and the maintenance of collateral, from some of the more ambitious players in the market.
Remarkably, this development was not encouraged by the supervisors. A report by a BIS committee representing the central banks of the Group of 10 admits that “The collateralization of inter-dealer exposures in principle could greatly reduce the likelihood that systemic disturbances are transmitted through that channel. However…the use of collateral entails other types of risk…With respect to liquidity risks, as the usage of collateral grows, dealers may become more vulnerable to liquidity pressures…[L]arge changes in market prices could produce sharp swings in the net values of the portfolios supported by collateral agreements which, in turn, could produce significant demands for collateral. Such collateral demands might prove especially difficult to meet in volatile markets, when meeting funds needs becomes more problematic in general.”
Yet that is, of course, the point of demanding collateralization and maintenance margins—to compel participants to recognize the liquidity risks in their ventures. There is, then, a task for supervision to judge whether in stress conditions these arrangements are sufficient to protect the system, at least until the fire hoses can be unloaded from the central banks? fire trucks.
The literature has failed to consider the extent to which market practitioners, with their great weight of money, influence both regulatory decisions and academic theory. Since the mid-1990s, the Basle committees have approved the use of internal value-at-risk analysis as the determinant of the amount of capital banks must hold to cover derivatives activities. This abdication of supervisory function followed JP Morgan?s publication of its RiskMetrics methodology, which supposedly enabled banks to measure the extent of their risks looking forward into the realm where uncertainty—not probability—reigns. In 1998 and 1999, the pressure from banks was to permit the use of Morgan?s CreditMetrics as a forecaster of the credit risk in a portfolio. The expertise of the banking industry is presumably in information-intensive lending, but CreditMetrics rests on published ratings of corporations and countries by Moody?s, Standard & Poor?s and Fitch/ICBA. It is not an accident, as ideologists like to say, that these operations proceed from Morgan. A former bank examiner from the New York Fed observes that in his time the approved approach to a problem was to go see what Morgan was doing and assume that was best practice. But that was before Dennis Weatherstone revised Morgan.
There is widespread agreement among banks and dealers that value-at-risk and RiskMetrics failed in third quarter 1998: Losses were far greater than the analysis had indicated was possible. Nevertheless, the banking supervisors have continued to recommend their use, faute de mieux. VAR is now to be supplemented by “stress tests,” a reasonable proposition—except that the banks have resisted suggestions from supervisors about the scenarios that should be tested to determine the stress resistance of the models. After all, they know what is plausible for their business, and the regulators are babes in these woods. But the usual argument that they have their own money at risk does not wash here, because most of the money at risk in banks is other people?s money.
The banks? attempt to gain approval for CreditMetrics as a measure by which capital could be allocated to credit risks was, fortunately, beaten back in spring 1999, though a more primitive use of ratings as guides to capital allocation was approved. William White, chief of research at BIS, has conceded that the use of ratings companies as monitors for the judgment of lending officers is an embarrassment to the industry and its supervisors, and will be revisited.
But the use of ratings as determinants of capital allocation is a truly awful idea. It creates a positive feedback loop, because the time when lending officers are too indulgent of their borrowers will coincide with the time when the ratings agencies are too generous with their ratings—and, worse, the time when lending officers get scared is the time when ratings agencies drop their ratings. A Wall Street veteran with significant government experience says, “The function of ratings agencies is to visit the battlefield after the fighting is over and shoot the wounded.” Moreover, formalizing the use of models like CreditMetrics would instantly create portfolio customs in the big banks that would encourage rebalancings to add lower-rated loans in times of upgrades and slough them in times of downgrades—even if the ratings for the specific loans in this portfolio had not changed. We already suffer enough—we don?t yet know how much—from brainlessness disguised as diversification, from the attempt to use black-box correlations as though we knew what caused them, from herd behavior induced by the feeding of common data into common programs.
As the “exchanges” become more and more electronic, the costs of operating in a public forum come down, but the big banks? prejudice against doing business in the open does not change. Since October 1998, Liffe has been offering what it describes as a “forward starting swap contract that cash settles on the start/effective date of the underlying swap. Consequent-ly, the contract captures the economic value of a swap without producing the actual swap cash flows, similar to a bond future capturing the economic value of a bond but without producing the actual coupon flows of that bond.” Two contracts are offered: for five years and for 10 years. The logic of offering the contract was that surpluses in U.S. budgets would lead to the repurchase of outstanding securities by the federal government, making the U.S. yield curve less useful as a benchmark, while the switch from Deutsche mark to euro denominations would give the German Bund less utility for the same purpose, especially as other countries in the European monetary union (and outside the EMU) become the dominant issuers of euro paper. In addition to its utility for traders who would wish to spread such contracts against Eurodollar and T-bill futures, the Liffe Euro-Financed Bond would thus offer a public service—and rival exchanges freely acknowledge it. But the contract has produced zero business, because it would deprive the banks that would be its dominant users of the information advantages they now have over their customers.
A good deal of today?s OTC derivatives trading expresses theories that explain nothing but itself. It is probably not desirable to perfect the cost-free interchange of pieces of paper issued in different countries under different legal systems for different purposes at different times. It is certainly not desirable to stretch a leveraged banking system to the point at which receipts and payments are so tightly articulated that a mere “displacement,” to use Hyman Minsky?s term, can trigger illiquidity. Although capital adequacy requirements and improved liquidity ratios are a step forward, there is insufficient recognition in our regulatory regime that it?s the weak creditor, not the weak debtor, who threatens the system—the inability of the borrower to pay matters systemically only to the extent that the lender cannot afford not to be repaid. Thus the Latin American failures of 1982-84 were a crisis in the United States, because Latin defaults would have devoured the capital of the American banking system, but the Thai/Korean/Indonesian ricochet was, as the President said, a bump in the road.
To look radically, at the roots, regulators should consider whether Modigliani-Miller?s and Robert Lucas? rational expectations theories may not be, simply, wrong. The truth is that assets financed with higher leverage do have greater volatility of values. And as Jack Hirshleifer argued almost 25 years ago, and Lynn Stout of Georgetown Law School has now pleaded, real life is better explained by theories of “heterogeneous expectations” (Hirshleifer uses the simpler “beliefs”) where market participants place their bets not only on the different information available to different participants but also, perhaps more importantly, on their differing reactions to the same information. In Stout?s precise formulation, “traders who share identical risk preferences and willingness to invest in information nevertheless may trade voluntarily in assets they neither produce nor consume if they make differing estimates of the probability distribution of future prices.” Although Hirshleifer specifically disavows Frank Knight?s distinction of uncertainty from risk, the case is even stronger if one rejects instead the modern alchemists who sell a purported talent for making the gold of risk from the dreck of uncertainty.
The argument against Hirshleifer has stressed that people with the wrong expectations would get weeded out, so that the only significant players would be those whose expectations were rational; but this assumes that banks learn from experience, and we know for sure that the half-life of such learning is quite short. Stout notes that a more likely reason for neglecting heterogeneous expectations is that they hugely complicate the task of model-making.
The major task of the supervisors in the next decade will be controlling and then reducing the role of correspondent banking in world finance. We have moved from the era of relationships to the era of transactions, without changing an institutional structure based in large part on the need to protect the privacy of relationships. Transparency requires execution through clearinghouses and exchanges, where smoke detectors can be installed to catch fires before they endanger the structure. Alfred Steinherr has suggested that regulators set capital requirements for OTC derivatives trades at a number that at least matches the margin requirements of the exchanges on which similar derivatives are traded. At this minimum, a significant number of banks might find the lower risk of the exchange-traded instrument sufficient inducement to move to the more transparent markets. If they don?t, one could raise the required capital allocation until they do. Concern about money-laundering should strengthen the right arms of the regulators: There are few more foolproof ways to launder money than the prearranged swap which gives the bank an intermediary?s riskless profit while funds are transferred from one side to the other.
Transparency has its admitted as well as its secret enemies. It may be true that if transactions cannot be kept out of sight for a while, some large deals will be doable only at a discount or a premium. But institutions should not be encouraged to believe that they can pick up or dispose of large quantities of securities without moving the price, because the day will certainly come when they can?t. Good brokers have never been willing to do business with big clients unless they know from the start how many shares are to be accumulated or sold; in the age of electronic brokerage, such information must be posted.
Transparency tells us NP, and, if extended to the revelation of large-trader positions, tells us also who is drunk and who is sober. Looking ahead, financial market supervisors must find a way to starve the appetite for leverage just when the party becomes most hungry. This may require the shutdown of offshore banking (not a bad idea in itself).
Derivatives markets guarantee a winner for every loser, but they will over time concentrate the losses in vulnerable sectors. Nature obeys Mayer?s Third Law, which holds that risk-shifting instruments will tend to shift risks onto those less able to bear them, because them as got want to keep and hedge while them as ain?t got want to get and speculate. The logic behind margin requirements in stock markets and capital requirements in banking also holds in the derivatives markets. Permitting highly leveraged institutions to hold private parties behind closed doors is the political version of selling volatility: the predictable likely gains will one day be overwhelmed by an equally predictable disastrous loss.
Commentary
Somebody Turn on the Lights
November 1, 1999