## Mr Greenspan discusses financial derivatives and the risks they entail
Speech by the Chairman of the Board of Governors of the Federal Reserve System, Greenspan, before the Futures Industry Association, Boca Raton, Florida, on 19 March
By far the most significant event in finance during the past decade has been the ext development and expansion of financial derivatives. This morning I should like to the scope of these markets, the nature of the risks they entail and some of the dif encounter in managing those risks.
At year-end, U.S. commercial banks, the leading players in global derivatives reported outstanding derivatives contracts with a notional value of $33 trillion, a has been growing at a compound annual rate of around 20 percent since 1990.
Of the $33 trillion outstanding at year-end, only $4 trillion were exchange-traded d the remainder were off-exchange or over-the-counter (OTC) derivatives. The greater OTC derivatives doubtless reflects the attractiveness of customized over stan products. But regulation is also a factor; the largest banks, in particular, seem regulation of exchange-traded derivatives, especially in the United States, as cre burdens than benefits. As I have noted previously, the fact that the OTC markets quite effectively without the benefits of the Commodity Exchange Act provides a argument for development of a less burdensome regime for exchange-traded finan derivatives.
Of course, notional values are not meaningful measures of the risks associat derivatives. Indeed, it makes no sense to talk about the market risk of derivative can be measured meaningfully only on an overall portfolio basis, taking into acco derivatives and cash market positions, and the offsets between them.
Clearly, the degree of counterparty credit risk on derivatives depends critically on which netting and margining procedures are employed to mitigate the risks. In the exchange-traded contracts, of course, daily variation settlements by clearing hous limit, if not totally eliminate, such counterparty risks.
In the case of OTC derivatives, counterparty credit exposures are far larger, though small fraction of the notional amounts. On a loan equivalent basis, a reasonably goo of such credit exposures, U.S. banks' counterparty exposures on such contracts are to have totaled about $325 billion last December. This amounted to less than 6 pe banks' total assets. Still, these credit exposures have been growing rapidly, more line with the growth of the notional amounts.
The leading role played by U.S. commercial and investment banks in the global derivatives markets is documented in a Bank for International Settlements survey of This survey estimated the size of the global OTC market at an aggregate notional val trillion, a figure that doubtless is closer to $80 trillion today. Once allowance i double-counting of transactions between dealers, U.S. commercial banks' share of thi market was about 25 percent, and U.S. investment banks accounted for another 15 pe While U.S. firms' 40 percent share exceeded that of dealers from any other country,
markets are truly global markets, with significant market shares held by dealers in Ca France, Germany, Japan, Switzerland and the United Kingdom.
Despite the world financial trauma of the past eighteen months, there is as yet no evide an overall slowdown in the pre-crisis derivative growth rates, either on or off excha Indeed, the notional value of derivatives contracts outstanding at U.S. commercial banks more than 30 percent last year, the most rapid annual growth since 1994. Although episo of extreme volatility have produced declines in the most highly leveraged contracts, growth of the more 'plain vanilla' products has continued apace or even accelerated.
The reason that growth has continued despite adversity, or perhaps because of it, is tha new financial instruments are an increasingly important vehicle for unbundling risks. T instruments enhance the ability to differentiate risk and allocate it to those investors and willing to take it. This unbundling improves the ability of the market to engender a product and asset prices far more calibrated to the value preferences of consumers than possible before derivative markets were developed. The product and asset price signals en entrepreneurs to finely allocate real capital facilities to produce those goods and servi valued by consumers, a process that has undoubtedly improved national productivity growt and standards of living.
Nonbank, as well as bank, users of these new financial instruments have increasing embraced them as an integral part of their capital risk allocation and profit maximizat should come as no surprise that the profitability of derivative products has been a major in the dramatic rise in large banks' non-interest earnings and doubtless is a factor significant gain in the overall finance industry's share of American corporate output d the past decade. In short, the value added of derivatives themselves derives from their to enhance the process of wealth creation.
While the value of risk unbundling has been known for decades, the ability to cre sophisticated instruments that could be effective in a dynamic market had to await the decade's development of computer and telecommunications technologies. The ability to crea and employ sophisticated financial products also galvanized the academic community t develop increasingly complex models of risk management. While recent history suggests th such models are useful, they are doubtless in need of much improvement - an issue to whic will return shortly.
Yet beneath all of the evidence of the value of derivatives to a market economy, there re a deep-seated fear that while individual risks seem clearly to have been reduced thr derivative-facilitated diversification, systemic risk has become enlarged, as a conseq Without question, derivatives facilitate the implementation of leveraged trading strat though the very technology that has made derivatives feasible has also improved the abili leverage without derivatives. Nonetheless, the possibility of increased systemic risk appear to be an issue that requires fuller understanding.
We should point out, first, the obvious. Overall, derivatives are mainly a zero sum game counterparty's market loss is the other counterparty's market gain. Counterparty cr exposures on OTC derivatives are a different issue and the source of much of the syste concerns. Such losses rose to record levels in the third quarter of 1998. Nonetheless, t of loss remained well below that on banks' loan portfolios. Moreover, the counterparty c
losses in the third quarter can be traced primarily to the extraordinary events in R produced many defaults on ruble forward contracts. In the fourth quarter such losses sharply, albeit not to the very low pre-crisis rate.
The bulk of the losses reported by the major derivative houses for the financiall third quarter of last year reflected declines in the market values of their under positions, especially in equities, commodities and emerging market debt. Der instruments were bystanders. They may well have intensified the losses in underlying but they were scarcely the major players.
Yet, through the past decades' phenomenal growth of the derivative market, there been a significant downturn in the economy overall that has tested the resilience of markets. (I operate on the premise that neither human nature nor the business cycle rendered obsolete.)
While nothing short of a major economic adjustment is likely to test the underlying of the derivative markets, there are reasons to believe that there are some fu strengths in these markets. First, despite the growing use of more exotic over-t instruments, the vast majority of trades are relatively straightforward interest rate swaps. The market risk on such swaps is presumably less daunting to indiv counterparties than their underlying exposures, or presumably the swaps would neve been initiated. Moreover, the credit risks are increasingly subject to comprehensi and margin requirements that, although they do not fully remove the risk, sign ameliorate it. And so far as banks are concerned, capital requirements are applied to as they are to loans that create credit risks quite similar to those of derivatives.
Hence, although one may harbor concerns about the overall capital adequacy of ban their degree of leverage, there is little to distinguish such concerns between risk and off-balance sheet claims.
The one area of risk that needs more thought is so-called potential future exposur particular point in time only a small fraction of the notional value of derivative in the money - that is, have a positive market value. Because prices will doubtless the future, those contracts with negative or even positive values have the potenti positive values and, hence, a potential credit loss on default.
That future potential for loss upon counterparty default will differ by the nat contract. For purposes of supervisory risk-based capital requirements, potenti exposure (over and above the current market value of derivatives, if positive) is estimated by separating derivatives into categories based on the underlying in (interest rate, exchange rate, commodity, equity, etc.) and the remaining maturity. requirement is then derived by applying fixed factors to each category that reflect in the price volatilities of the instruments and the structure of the contracts. Int (70 percent of the notional value of OTC derivatives) have limited long-term loss primarily because the contracts do not provide for an exchange of principal and the is effectively amortized as interest payments are exchanged over the life of the Foreign exchange, commodity and equity derivatives, of course, entail far greater e either because principal amounts are exchanged or because the underlying's price volatile.
This approach to regulatory capital requirements is not altogether satisfactory. The sophisticated derivative dealers parse their derivatives book in more detail. And certa single point estimate cannot capture the range of losses that might reasonably be experie Hence, in evaluating derivatives risk, far more stress testing of the lower probability o is a necessity. Even a one in 500 occurrence does happen once every 500 times, and if occurrence could threaten the franchise value of the derivatives counterparty it is an im concern for risk aversion.
But we have to be careful of how we view these ostensibly low-probability events. They low-probability only if we presume that the reality from which these events derive is represented by a single bell-shaped probability distribution, be it a normal distribution a fat-tailed one.
Modern quantitative approaches to risk measurement and risk management take as their starting-point historical experience with market price fluctuations, which is stati summarized in probability distributions. We live in what is mostly a stable economic sys in which market imbalances give rise to continuous and inevitable moves toward equilibri resolutions. However, the violence of the responses to what seemed to be relatively imbalances in southeast Asia in 1997 and throughout the global economy in August and September of 1998 have raised the possibility of a discontinuous adjustment process.
Almost all the time investors adopt strategies that seek profit only in a relatively l context, fostering the propensity for convergence toward equilibrium that ordinar characterizes financial markets. But from time to time (and quite possibly with incre frequency) the resulting propensity toward convergent equilibrium has given way as invest suffer an abrupt collapse of comprehension of, and confidence in, future economic even Risk aversion accordingly rises dramatically and deliberative trading strategies are repl rising fear-induced disengagement. Yield spreads on relatively risky assets wid dramatically. In the more extreme manifestation, the inability to differentiate among de of risk drives trading strategies to ever more liquid instruments. Strategies become so t that traders want the capacity to reverse decisions at minimum cost. As a consequence, among riskless assets, illiquidity premiums rise dramatically as investors seek the h traded 'on-the-run' issues.
History tells us that sharp reversals in confidence happen abruptly, most often with advance notice. They are self-reinforcing processes that can compress into a very short period. Panic market reactions are characterized by a dramatic shift to maximize shortvalue, and are an extension of human behavior that manifests itself in all forms of h interaction - a set of responses that does not seem to have changed over the generatio defy anyone to distinguish a speculative price pattern for 1999 from one for 1899 if the specify neither the dates nor the levels of the prices.
If this paradigm turns out to be the appropriate representation of the way our economy an financial markets will work in the future, it has significant implications for risk mana Probability distributions estimated largely, or exclusively, over cycles excluding peri panic will underestimate the probability of extreme price movements because they fail capture a secondary peak at the extreme negative tail that reflects the probability of oc of a panic. Furthermore, joint distributions estimated over panicless periods
underestimate the degree of correlation between asset returns during panics when disengagement by investors result in simultaneous declines (or, in rare instances, i values as investors no longer adequately differentiate among degrees of risk and Consequently, the benefits of portfolio diversification will tend to be si overestimated by current models.
Such a view of the world would also have important implications for approaches prudential oversight of capital adequacy for banks and other financial institutions. minimum capital requirements for banks' trading portfolios are now based on the bank internal risk measurement models. Furthermore, regulators are exploring the poten using an internal models approach to credit risk in the banking book.
Some may now argue that the periodic emergence of financial panics implies a ne abandon models-based approaches to regulatory capital and to return to traditional a based on regulatory risk measurement schemes. In my view, however, this would be a mistake. Regulatory risk measurement schemes are simpler and much less accurate banks' risk measurement models. Consequently, they provide banks with the motive an opportunity to engage in regulatory arbitrage that seriously undermines the regulator and frustrates the underlying safety and soundness objective. Specifically, they in to reduce holdings of assets where risks and regulatory capital are overestimated by and increase holdings of assets where risks are underestimated by regulators.
It would be far better to provide incentives for banks to enhance their risk procedures by taking account of the potential existence and implications of disco episodes. Scenario analysis can highlight vulnerabilities to the kind of flights t flights to liquidity that seem increasingly frequent. Stress testing of correlatio can reveal the disappearance of apparent diversification benefits in such scenarios.
Stress testing requirements already are part of the internal models approach t requirements for market risks in bank trading accounts. Stress testing of est counterparty credit risks should also be required. The logic is the same as for mark factors that are used to determine supervisory capital requirements for counterpa exposures are based on statistical analyses of non-panic periods. Moreover, duri periods the usual assumption that potential future exposures are uncorrelated wit probabilities becomes invalid. For example, the collapse of emerging market currenc greatly increase the probability of defaults by residents of those countries at the s exposures on swaps in which those residents are obligated to pay foreign curre increasing dramatically.
Supervisors should avoid any temptation to increase the supervisory factors for future exposure to address these crisis scenarios, which have vastly different impl different combinations of contracts and counterparties. But they can and should re requirements relating to the scenarios to be simulated by the bank and the incorpo stress test results into the policies and limits set by the bank's management an directors.
As we approach the twenty-first century, both banks and nonbanks will need to cont reassess whether their risk management practices have kept pace with their own e activities and with changes in financial market dynamics and readjust accordingly.
they succeed I am quite confident that market participants will continue to increase reliance on derivatives to unbundle risks and thereby enhance the process of wealth creat