Chapter

II. Background Papers: Financial Supervisory and Regulatory Issues

Author(s):
International Monetary Fund
Published Date:
January 1995
Share
  • ShareShare
Show Summary Details

I Capital Adequacy and Internal Risk Management

Banks now actively participate in derivative markets, especially in the largely unregulated over-the-counter (OTC) derivative markets. They also now operate in a global financial environment in which new products, new pricing techniques, and new risk-management techniques are being introduced continuously and at breakneck speed. During the last few years, regulators and leaders in the financial industry have been engaged in a process of designing a new, more relevant set of regulatory capital requirements for banks that operate in this high-risk and fast-changing environment. In effect, structural changes in financial markets led regulators to reconsider the traditional approaches to supervision and regulation and to develop new more flexible approaches to bank regulation. Through this process, regulators have come to recognize private internal risk management as the foundation for ensuring the safety and soundness of individual financial institutions, and in reducing systemic risk.

Although this process is likely to continue, bank regulators are now considering revolutionary changes in the way banks are supervised and regulated. On April 12, 1995, the Basle Committee on Banking Supervision issued for comment a new set of capital requirements that allows banks to use their own internal risk-management models to estimate an important component for determining regulatory capital requirements—value-at-risk. The proposal moves bank regulation in a new direction in that it allows banks to actively participate in the design of the framework for establishing capital requirements. The Basle Committee expects comments by July 1995. The plan is to adopt the new requirements by the end of the year and have them implemented by the end of 1997.

A review of the developments that led regulators to take this new direction is followed by an examination of the new Basle approach for calculating capital requirements. This is followed by a discussion of current efforts to improve internal risk management by banks, which is likely to become an integral part of the internal model-based approach to market risk-based capital requirements.

Background

Growth and Impact of Derivatives

Between the end of 1986 and the end of 1994, the total notional principal of outstanding exchange-traded derivative contracts—including interest rate futures and options, currency futures and options, and stock-market index futures and options—has grown at an annual average rate of 140 percent, from $0.6 trillion to $8.8 trillion (Table II.1). In the same period, annual turnover has grown about four times, from 315 million contracts to 1,140 million contracts a year (Table II.2).

Table II.1.Markets for Selected Derivative Financial Instruments: Notional Principal Amounts Outstanding(In billions of U.S. dollars, end-year data)
198619871988198919901991199219931994
Interest rate futures370.0487.7895.41,200.81,454.52,156.72,913.04,942.65,757.4
Futures on short-term interest rate instruments274.3338.9721.71,002.61,271.11,906.32,663.74,616.75,401.8
Three-month Eurodollar1229.5307.8588.8671.9662.61,100.51,389.62,178.72,468.6
Three-month Euro-yen2109.5243.5254.5431.81,080.11,467.4
Three-month Euro-deutsche mark314.447.7110.0229.2421.9425.7
Futures on long-term interest rate instruments95.7148.8173.7198.2183.4250.4249.3325.9355.6
U.S. Treasury bond423.026.539.933.223.029.831.332.636.1
Notional French Government bond52.17.67.06.17.011.421.012.612.7
Ten-year Japanese Government bond663.5104.8106.7129.5112.9122.1106.1135.9164.3
German Government bond71.44.213.720.227.833.341.7
Interest rate options8146.5122.6279.2387.9599.51,072.61,385.42,362.52,622.7
Currency futures10.214.612.115.916.917.924.932.233.0
Currency options839.259.548.050.256.562.870.975.454.5
Stock market index futures14.517.827.141.369.176.079.7109.9127.7
Stock market index options837.827.742.970.693.7132.8158.6238.3242.4
Total618.3729.91,304.81,766.62,290.23,518.84,632.57,760.88,837.8
In the United States517.9577.7950.31,152.31,263.02,132.12,675.54,324.94,754.9
In Europe13.113.3177.7250.8461.2710.11,114.31,777.91,832.0
In Japan63.5107.7106.6260.9424.2441.2576.11,193.51,498.2
In other countries23.731.270.2102.6141.8235.4266.6464.5752.8
Source: Bank for International Settlements.

Traded on the Chicago Mercantile Exchange-International Monetary Market (CME-IMM), Singapore Mercantile Exchange (SIMEX), London International Financial Futures Exchange (LIFFE), Tokyo International Financial Futures Exchange (TIFFE), and Sydney Futures Exchange (SFE).

Traded on the TIFFE and SIMEX.

Traded on the Marché à Terme International de France (MATIF) and LIFFE.

Traded on the Chicago Board of Trade (CBOT), LIFFE, Mid-America Commodity Exchange (MIDAM), New York Futures Exchange (NYFE), and Tokyo Stock Exchange (TSE).

Traded on the MATIF.

Traded on the TSE, LIFFE, and CBOT.

Traded on the LIFFE and the Deutsche Terminborse (DTB).

Calls plus puts.

Source: Bank for International Settlements.

Traded on the Chicago Mercantile Exchange-International Monetary Market (CME-IMM), Singapore Mercantile Exchange (SIMEX), London International Financial Futures Exchange (LIFFE), Tokyo International Financial Futures Exchange (TIFFE), and Sydney Futures Exchange (SFE).

Traded on the TIFFE and SIMEX.

Traded on the Marché à Terme International de France (MATIF) and LIFFE.

Traded on the Chicago Board of Trade (CBOT), LIFFE, Mid-America Commodity Exchange (MIDAM), New York Futures Exchange (NYFE), and Tokyo Stock Exchange (TSE).

Traded on the MATIF.

Traded on the TSE, LIFFE, and CBOT.

Traded on the LIFFE and the Deutsche Terminborse (DTB).

Calls plus puts.

Table II.2.Annual Turnover in Derivative Financial Instruments Traded on Organized Exchanges Worldwide1(In millions of contracts traded)
198619871988198919901991199219931994
Interest rate futures91.0145.7156.3201.0219.1230.9330.1427.1627.8
Futures on short-term interest rate instruments16.429.433.770.276.084.8130.8161.0252.9
Three-month Eurodollar112.423.725.246.839.441.766.970.2113.6
Three-month Euro-yen24.715.216.217.426.944.2
Three-month Euro-deutsche mark31.63.14.812.221.429.5
Futures on long-term interest rate instruments74.6116.3122.6130.8143.1146.1199.3266.1374.9
U.S. Treasury bond454.669.473.872.878.269.971.780.7101.5
Notional French Government bond51.111.912.415.016.021.131.136.850.2
Ten-year Japanese Government bond69.418.418.919.116.412.912.115.614.1
German Government bond70.35.39.612.418.927.751.5
Interest rate options822.329.330.539.552.050.864.882.9114.5
Currency futures19.921.222.528.229.730.031.339.069.7
Currency options813.018.318.220.718.922.923.423.821.3
Stock market index futures28.436.129.630.139.454.652.071.2109.0
Stock market index options8140.4139.179.1101.7119.1121.4133.9144.1197.9
Total315.0389.6336.2421.2478.3510.5635.6788.01,140.2
In the United States288.4317.6251.4287.0310.9301.5340.1380.3509.5
In Europe10.335.940.764.483.0110.5185.0263.5398.5
In Japan9.418.523.145.760.666.251.757.870.5
In other countries6.917.721.024.123.832.458.886.4161.7
Source: Bank for International Settlements.

Traded on the Chicago Mercantile Exchange-International Monetary Market (CME-IMM), Singapore Mercantile Exchange (SIMEX), London International Financial Futures Exchange (LIFFE), Tokyo International Financial Futures Exchange (TIFFE), and Sydney Futures Exchange (SFE).

Traded on the TIFFE and SIMEX.

Traded on the Marché à Terme International de France (MATIF) and LIFFE.

Traded on the Chicago Board of Trade (CBOT), LIFFE, Mid-America Commodity Exchange (MIDAM), New York Futures Exchange (NYFE), and Tokyo Stock Exchange (TSE).

Traded on the MATIF.

Traded on the TSE, LIFFE, and CBOT.

Traded on the LIFFE and the Deutsche Terminbörse (DTB).

Calls plus puts.

Source: Bank for International Settlements.

Traded on the Chicago Mercantile Exchange-International Monetary Market (CME-IMM), Singapore Mercantile Exchange (SIMEX), London International Financial Futures Exchange (LIFFE), Tokyo International Financial Futures Exchange (TIFFE), and Sydney Futures Exchange (SFE).

Traded on the TIFFE and SIMEX.

Traded on the Marché à Terme International de France (MATIF) and LIFFE.

Traded on the Chicago Board of Trade (CBOT), LIFFE, Mid-America Commodity Exchange (MIDAM), New York Futures Exchange (NYFE), and Tokyo Stock Exchange (TSE).

Traded on the MATIF.

Traded on the TSE, LIFFE, and CBOT.

Traded on the LIFFE and the Deutsche Terminbörse (DTB).

Calls plus puts.

Meanwhile, OTC derivative markets have expanded at a similar pace. The notional principal of outstanding interest rate and currency swaps increased from $1.0 trillion at the end of 1987 to almost $8.0 trillion at the end of 1993 (Table II.3). In the second half of 1993 alone, close to $2.4 trillion new interest rate swaps and currency swaps were transacted (Table II.4). In addition, the total notional principal of more complex swap-related OTC derivative products—caps, collars, floors, and swaptions—has increased from $0.6 trillion at the end of 1990 to $1.4 trillion at the end of 1993. Underlying this rapid growth in OTC derivative markets has been the increased globalization of derivative activities. At the end of 1987, non-U.S. dollar-denominated interest rate swaps represented about 21 percent of all interest rate swaps; by the end of 1992, this share had expanded to more than 60 percent (Table II.5).

Table II.3.Notional Principal Value of Outstanding Interest Rate and Currency Swaps1(In billions of U.S. dollars)
1987198819891990199119921993
Interest rate swaps
All counterparties682.91,010.21,502.62,311.53,065.13,850.86,177.3
Interbank (ISDA member)206.6341.3547.1909.51,342.31,880.82,967.9
Other (end-user and brokered)476.2668.9955.51,402.01,722.81,970.13,209.4
End-user476.2668.9955.51,402.01,722.81,970.13,209.4
Financial institutions300.0421.3579.2817.1985.71,061.11,715.7
Governments247.663.276.2136.9165.5242.8327.1
Corporations3128.6168.9295.2447.9571.7666.21,166.6
Unallocated15.54.9
Brokered
Currency swaps
All counterparties365.6639.1898.21,155.11,614.31,720.81,799.2
(Adjusted for reporting of both sides)(182.8)(319.6)(449.1)(577.5)(807.2)(860.4)(899.6)
Interbank (ISDA member)71.0165.2230.1310.1449.8477.7437.0
Other (end-user and brokered)294.6473.9668.1844.91,164.61,243.11,362.2
End-user4147.3237.0334.1422.5582.3621.5681.1
Financial institutions61.9102.7141.7148.2246.7228.7221.9
Governments233.954.065.683.296.9110.6135.8
Corporations351.676.5116.5191.1238.7282.2323.4
Unallocated3.810.3
Brokered
Total (interest rate and currency swaps for all counterparties)1,048.51,649.32,400.83.466.64.679.45,571.67.976.5
Sources: Bank for International Settlements, International Banking and Financial Market Developments, various issues; and International Swaps and Derivatives Association, Inc. (ISDA).

As of end of December.

Including international institutions.

Including others.

Adjusted for double-counting as each currency swap involves two currencies.

Sources: Bank for International Settlements, International Banking and Financial Market Developments, various issues; and International Swaps and Derivatives Association, Inc. (ISDA).

As of end of December.

Including international institutions.

Including others.

Adjusted for double-counting as each currency swap involves two currencies.

Table II.4.New Interest Rate and Currency Swaps1(In billions of U.S. dollars)
1987198819891990199119921993
First

half
Second

half
First

half
Second

half
First

half
Second

half
First

half
Second

half
First

half
Second

half
First

half
Second

half
First

half
Second

half
Interest rate swaps
All counterparties181.5206.3250.5317.6389.2444.4561.5702.8762.1859.71,318.31,504.31,938.52,166.3
Interbank (ISDA member)58.967.086.6106.5140.4177.6223.2261.3335.4426.4617.7718.7959.21,044.8
Other (end-user and brokered)122.6139.3163.9211.1248.7266.8338.2441.5426.8433.3700.6785.6979.31,121.5
End-user121.0136.0162.3209.1242.8260.6334.5370.8419.2425.5681.0755.7922.91,077.7
Financial institutions82.386.4102.8135.3152.9165.0200.2219.9229.3263.1404.6449.3518.1597.7
Governments210.910.815.717.223.016.633.741.043.435.564.984.0107.790.8
Corporations327.834.843.954.360.579.0100.6110.0146.4126.9211.5222.4288.8389.2
Unallocated4.12.36.58.3
Brokered1.63.31.61.95.96.23.770.77.67.719.629.956.543.8
Currency swaps
All counterparties87.185.7122.3126.2166.7189.6189.3236.2322.6334.1312.1291.6313.6276.7
(Adjusted for reporting of both sides)(43.5)(42.8)(61.1)(63.1)(83.4)(94.8)(94.6)(118.1)(161.3)(167.1)(156.1)(145.8)(156.8)(138.4)
Interbank (ISDA member)17.518.325.433.350.850.553.069.6105.9102.068.364.261.349.6
Other (end-user and brokered)69.567.496.992.9115.9139.1136.3166.6216.7232.1243.9227.4252.3227.2
End-user434.333.547.546.457.569.667.783.0103.1116.0121.6113.1126.0112.9
Financial institutions18.913.023.320.222.429.822.828.641.157.440.938.039.337.9
Governments27.66.310.68.713.29.812.510.913.717.123.618.530.821.8
Corporations37.913.612.916.218.527.732.443.548.241.557.156.655.953.1
Unallocated0.60.71.33.52.20.1
Brokered0.90.31.90.21.00.90.710.70.10.71.20.31.3
Total (interest rate and currency swaps for all counterparties)268.6292.0372.8443.8555.9633.9750.8939.01,084.71,193.81,630.41,795.92,252.12,443.0
Sources: Bank for International Settlements, International Banking and Financial Market Developments, various issues; and International Swaps and Derivatives Association, Inc. (ISDA).

During the respective half of the year.

Including international institutions.

Including others.

Adjusted for double-counting as each currency swap involves two currencies.

Sources: Bank for International Settlements, International Banking and Financial Market Developments, various issues; and International Swaps and Derivatives Association, Inc. (ISDA).

During the respective half of the year.

Including international institutions.

Including others.

Adjusted for double-counting as each currency swap involves two currencies.

Table II.5.Currency Composition of Notional Principal Value of Outstanding Interest Rate and Currency Swaps(In billions of U.S. dollars)
1987198819891990199119921993
Interest rate swaps
All counterparties682.91,010.21,502.62,311.53,065.13,850.86,177.3
U.S. dollar541.5728.2993.71,272.71,506.01,760.22,457.0
Japanese yen40.578.5128.0231.9478.9706.01,247.4
Deutsche mark31.656.584.6193.4263.4344.4629.7
Pound sterling29.752.3100.4242.1253.5294.8437.1
Other39.594.8195.8371.5563.3745.31,406.1
Interbank (ISDA member)206.6341.3547.1909.51,342.31,880.82,967.9
U.S. dollar161.6243.9371.1492.8675.0853.91,008.4
Japanese yen19.543.061.1126.1264.9441.3820.8
Deutsche mark7.917.232.678.4111.2175.6356.1
Pound sterling10.417.640.0100.1106.3137.2215.2
Other7.119.642.2112.1184.9272.7567.4
End-user476.2668.9955.51,402.01,722.81,970.12,209.4
U.S. dollar379.9484.3622.6779.9831.0906.31,448.6
Japanese yen21.035.566.9105.8214.0264.7426.7
Deutsche mark23.739.352.0115.0152.2168.8273.7
Pound sterling19.334.760.4142.0147.3157.6222.0
Other32.475.2153.6259.4378.3472.7838.4
Currency swaps1
All counterparties182.8319.6449.1577.5807.2860.4899.6
U.S. dollar81.3134.7177.1214.2292.1309.0320.0
Japanese yen29.965.5100.6122.4180.1154.3158.8
Deutsche mark10.717.026.936.247.653.369.7
Pound sterling5.38.916.724.537.440.144.2
Other55.793.5127.8180.3250.0303.6306.9
Interbank (ISDA member)35.582.6115.1155.1224.9238.9218.5
U.S. dollar16.734.148.259.786.890.982.3
Japanese yen7.218.628.337.460.953.953.2
Deutsche mark1.63.05.47.69.412.612.9
Pound sterling1.11.64.36.28.410.47.1
Other9.025.428.844.159.571.163.0
End-user147.3237.0334.1422.5582.3621.5681.1
U.S. dollar64.6100.7128.9154.5205.3218.1237.7
Japanese yen22.747.072.285.0119.2100.4105.6
Deutsche mark9.114.021.528.538.240.756.8
Pound sterling4.27.312.418.329.029.737.0
Other46.768.199.0136.2190.6232.6244.0
Sources: Bank for International Settlements, International Banking and Financial Market Developments, various issues; and International Swaps and Derivatives Association, Inc. (ISDA).

Adjusted for double-counting as each currency swap involves two currencies.

Sources: Bank for International Settlements, International Banking and Financial Market Developments, various issues; and International Swaps and Derivatives Association, Inc. (ISDA).

Adjusted for double-counting as each currency swap involves two currencies.

Accompanying this overall expansion in derivative markets has been a gradual, but now very noticeable, change in the nature of banking, in which the activity of banks has shifted from traditional lending to derivative markets activities. Although these changes have occurred across a wide range of countries, they are seen most clearly in the activities of large U.S. banks. Among the seven major U.S. money center banks, interest income has declined, on average, from 70 percent of total revenue at the end of 1987 to less than 50 percent at the end of 1993. By contrast, trading income has more than doubled from 5⅔ percent of total revenue to more than 13½ percent and fee income revenue has increased from about 12½ percent to about 17¾ percent. By the end of 1992, the notional principal of interest rate swaps held off-balance sheet by the seven largest U.S. money center banks ($1.7 trillion) was almost twice as large as their total balance sheet assets ($854 billion), reflecting the increasing concentration of bank activities in the OTC derivative transactions.

Advances in technology, the rapid growth of the OTC derivative markets, the globalization and liberalization of financial markets, and the change in bank activities have the following important implications: (1) market and operational risk have become more important for banks;1 (2) new and more complicated products are being introduced continuously; (3) linkages between financial markets and institutions have strengthened significantly; (4) banks now have the ability to change their portfolios and risk exposures very quickly; and (5) financial market conditions and the value of bank portfolios can change very rapidly. These structural changes in the financial services industry have forced bank supervisors and regulators and other financial market regulators to re-examine the current regulatory regime.

Earlier Proposals for Revising Capital Adequacy Standards

Until recently, the most important tools of bank supervisors and regulators have been capital requirements, on-site examinations, and regular reporting requirements. As discussed in previous capital markets reports, shortly after the 1988 Basle Accord was agreed, both supervisors and bankers came to see that the minimum capital adequacy standards defined in the Accord were inadequate, in part because of changes in the nature of banking and, more generally, in the financial services industry.2 It was recognized that the Basle Accord ignored market risk, by focusing primarily on credit risk, and did not cover adequately many new and complex OTC derivative products. The Accord was drafted at a time when most of these products did not exist, when banks were concentrating on traditional banking functions, and when credit risk was the single most important risk factor.3

Despite these weaknesses, the minimum capital adequacy standard defined in the Accord (slightly modified in July 1994 to accommodate more general recognition of bilateral netting arrangements) is still the current international standard for bank capital requirements.4 As a result of these perceived weaknesses, the European Union (EU) released in March 1993 a new Capital Adequacy Directive (CAD) for implementation starting January 1996.5 Similarly, the Basle Committee on Banking Supervision issued for comment, in April 1993, a proposal on a capital standard based on market risk.6 Under both the EU and Basle initiatives, a bank’s assets and liabilities would be separated into a banking book, which includes traditional bank loans and deposits, and a trading book, which includes short-term trading and hedging positions. Capital charges for positions in the banking book would continue to be determined according to the 1988 Basle Accord. However, the capital charges for positions in the trading book would be determined by a new set of market-risk-based capital standards that utilize a product-based approach. There are separate rules for three major classes of products—equity, interest rate, and exchange rate products. A building block approach, by which market risk and specific risk would be considered separately, would be used to compute capital charges for each product type, which would then be summed to generate capital requirement for the trading book.

Weaknesses of the 1993 Basle Proposal and the CAD

The Basle and EU proposals have stimulated intense discussions among regulators and market participants on the appropriate approach for determining market-risk-based capital requirements. Market participants have suggested that both proposals are too complex for small banks and for the public to understand and too crude for banks active in the derivative markets. It also has been suggested that the proposals, if adopted, are likely to create inefficiencies because banks would have to maintain two risk-measurement systems: one for calculating regulatory capital requirements, and one for daily risk management and trading. The proposals would also reduce banks’ incentives to improve their own risk-management system.

The Basle proposal has also been challenged on the grounds that it cannot yield accurate capital requirements, not because it lacks rigor or complexity but because it employs ad hoc disallowances to account for spread risks and gap risks;7 it uses crude maturity bands and sensitivity factors to account for interest rate risk; and it ignores correlations and offsets between risk factors by simply adding the capital charges for different classes of products together to give the aggregate capital requirement. This last feature represents a conservative approach because it is justified only if all product classes are perfectly correlated. Some have argued that this approach might not reduce risk in the long run; while a bank still has the flexibility to increase its capital reserves when its capital requirement is too low, it cannot reduce its capital requirement when it is too high. Excessive capital requirements can reduce the competitiveness of banks relative to nonbanks and, in the long run, affect the banking system.

Another concern often expressed about the two initiatives is that uneven capital charges might distort bank activities. For example, because both proposals ignore correlations among product classes, hedging and diversification will be discouraged because neither will lead to lower capital requirements. In addition, the Basle Accord, the Basle proposal, and the CAD ignore the risk of a change in the sensitivity of derivative prices to the price of the underlying primary asset (gamma risk) and the risk of a change in the volatility of the underlying asset (vega risk). The proposed capital standards would, therefore, encourage options strategies, such as straddles (bets on changes in volatility).8

Basle’s product-based approach for determining capital charges has been criticized as being unduly rigid. Given demand pressures and the rapid pace of advances in technology, new hybrid products are being introduced continuously. The Basle proposal does not have the flexibility to cover these new products. Others are also uncomfortable with applying market-risk-based capital requirements to the trading book and credit-risk-based capital requirements to the banking book, because items in the banking book entail market risk. Furthermore, market risk and credit risk are not unrelated. A default on a derivative contract can occur because of a major market movement. In addition, by classifying items into banking and trading books according to the length of the intended holding period, significant distortions in banking activities might be created. In effect, it has been argued that it might be more appropriate to apply the same standard on the entire portfolio of a bank. Finally, capital requirements are based on a snapshot of risk exposures. When dynamic trading strategies are employed, however, the EU and Basle approaches will not accurately measure the risk inherent in these strategies.9

The intense discussions surrounding the Basle and EU initiatives have clarified that from the perspective of the financial industry, an “ideal” capital standard would have the following characteristics: (1) simple enough to be understood and implemented by even the less sophisticated banks; (2) as close as possible to a bank’s internal risk-management system; (3) comprehensive enough to avoid distorting bank activities; (4) built on a full portfolio approach, taking into account correlations and offsets; (5) easily extendable to cover new products and new markets; (6) applicable to all banks regardless of their size and primary activities; and (7) structured in a way such that improvements in internal risk-management systems are encouraged.

An important question is whether it is possible for regulators to establish a set of capital adequacy standards with these characteristics. The intense discussions among regulators and market participants, following the Basle proposal and the CAD, have led to the conclusion that the traditional approach established in the 1988 Basle Accord—and extended by the Basle proposals and the CAD—might not work very well in a dynamic environment where product innovations can render obsolete any rigidly defined minimum capital adequacy standard.

Basle Committee’s New Approach: Using Internal Risk-Management Models

In light of the difficulties encountered in producing a new set of comprehensive capital adequacy requirements, the Group of Ten regulators have apparently taken a new direction toward the calculation of capital adequacy requirements. On April 12, 1995, the Basle Committee on Banking Supervision issued a proposal that recommends a new approach toward the calculation of bank capital requirements. The new approach would allow banks, for the first time, to use their internal risk-management models to determine regulatory capital requirements. Instead of adhering to a detailed framework for computing risk exposures (for reporting purposes) and capital requirements, banks would be able, under certain conditions, to use their own models—the ones they use for day-to-day trading and risk management—to determine an important component of their regulatory capital requirements.

Value-at-Risk: Key to Risk Measurement

Although Basle’s new approach for establishing capital adequacy requirements does not, as proposed, restrict a bank from using one or another risk-measurement model or estimation method, it does require that banks use a common measure of risk. Specifically, the Basle Committee advocates “value-at-risk” as the standard measure for risk exposures.10 Value-at-risk is an estimate of the “maximum” loss in the value of a portfolio or financial position over a given time period with a certain level of confidence. This level of confidence is represented by the probability that the actual loss will not exceed a prespecified “maximum.” The probability is usually referred to as the confidence interval. The new approach requires the use of a value-at-risk corresponding to a ten-day holding period and a 99 percent confidence interval. Under these standards, a $100 million value-at-risk means that there is a 99 percent chance that the loss in the portfolio value over a ten-day period will be less than $100 million. Specifying value-at-risk as the standard risk measure is not more than specifying a common unit of measurement.

Value-at-risk can be, and currently is, estimated by using various techniques. The so-called asset-normal approach is commonly used and assumes that asset returns are jointly normally distributed.11 Under this assumption, the probability distribution of the rate of return of a portfolio can be estimated easily and quickly, which makes the calculation of value-at-risk relatively easy. The main disadvantage of this approach is that the value-at-risk calculation might not be accurate if the normal distribution does not offer a good description of the underlying price data. Some have argued that the t-distribution, which allows extreme price changes to occur more often, offers a better description of the statistical properties of asset returns. Furthermore, the asset-normal approach cannot easily deal with option risk.12

An approach that can be used to compute the value-at-risk of a portfolio with options and other derivative instruments is the so-called delta normal method. This approach, which is used in Basle’s 1993 proposal and the CAD, treats an option as a position equal to the market value of the underlying asset multiplied by the delta13 of the option.14 This approach is conceptually and computationally easy, but it ignores the effect of a change in the delta of an option (the gamma risk).15 More importantly, since the linear approximation only works for small changes in the process of the underlying asset prices, the approach can produce a significantly biased estimate of value-at-risk when large price changes are being considered.

The “delta gamma” approach is an extension of the delta normal approach. It incorporates both the delta and the gamma of an option to construct an approximation of the option position.16 While this approach can yield more accurate results, it is also computationally more intensive. In the spirit of this approach, Basle’s new approach also contains a provision for gamma and vega risk under the heading of a “delta-plus” method for banks that do not want to follow the internal model-based approach for calculating their regulatory capital requirements.17

An approach that has become very popular among the sophisticated derivative houses is the so-called model-simulation or Monte Carlo approach. This approach directly assesses the value of a derivative asset by evaluating its price for a large number of simulated price paths for the underlying asset, which are generated according to a particular dynamic model. An advantage of this approach is that it can be used for any kind of derivative contract. Furthermore, provided that the assumed price dynamics are correct, this method would produce more accurate estimates of value-at-risk because it does not rely on approximations of any kind. This method, however, introduces model risk into the calculation—the risk that the underlying model of price dynamics might not be correct. Moreover, because banks make different assumptions about price behavior, the same Monte Carlo approach can produce very different estimates of value-at-risk.

There are also many other alternative approaches that can be used to estimate value-at-risk: the historical distribution method, which utilizes the past distribution of asset returns;18 the so-called factor-push approach, which computes value-at-risk by considering hypothetical changes in the value of the underlying risk factors;19 and the maximum loss approach, which computes value-at-risk by locating the combination of asset prices that can yield the biggest loss for a given probability.20

To summarize, the new 1995 Basle proposal does not impose restrictions on the choice of models or the estimation method a bank can use to calculate value-at-risk. Even for the simplest asset-normal approach, there are various ways to estimate the asset return standard deviations and correlations. One can employ standard statistical formulas for these statistics. Alternatively, some have suggested that older observations can be assigned a smaller weight in the estimation;21 others have suggested that such weights should be estimated simultaneously with the statistics. These differences in estimation methods can lead to differences in estimates of value-at-risk for the same portfolio (Box II.2).22

Capital Requirements for Banks Using Internal Models

Under Basle’s new approach, a bank’s capital requirement on a particular day would be the higher of either the previous day’s value-at-risk or the average of value-at-risk on each of the preceding sixty business days, multiplied by a safety factor that is greater than or equal to three. The safety factor creates a buffer to safeguard against the underestimation of a bank’s capital requirement. The exact value of the safety factor would be related to the ex post performance of the bank’s internal model; that is, a bank with a model that produces estimates of value-at-risk that are regularly inaccurate would be obliged to use a higher safety factor. The proposal does not explicitly specify how this safety factor should be determined in cases when it would exceed three, however. In addition to the capital charges related to value-at-risk, a bank that uses its internal model would be subject to a separate capital charge that covers the specific risk of traded debt and equity securities if such risk has not been incorporated into the model.

Quantitative and Qualitative Standards

In addition to requiring the use of a ten-day value-at-risk corresponding to a 99 percent confidence interval, the 1995 Basle proposal also contains a number of other requirements on the computation of the value-at-risk. For instance, the sample period for estimating the value-at-risk is required to cover at least a one-year period. The data set is also required to be updated at least once every three months. Furthermore, estimates of value-at-risk from different risk categories must be aggregated by simple summing, although the use of correlations between assets in the same risk category is allowed in the estimation of value-at-risk for a particular risk category.

The new approach also contains special requirements on the measurement of option risk. First, a minimum ten-day holding period is to be applied to option positions. Second, because option risk is nonlinear, banks are not allowed to estimate the ten-day value-at-risk by scaling up the estimate of daily value-at-risk by the square root of time. Third, banks are required to measure the volatilities of option positions according to their maturities.

In addition to quantitative standards, banks that choose to use their internal models to determine market-risk-based capital requirements are required to meet several qualitative standards. These standards specify minimum requirements on a bank’s internal control and risk-management process. First, the bank is required to maintain an independent risk-control unit that reports directly to senior management. Second, the unit must conduct ex post performance analysis regularly to check the accuracy of the bank’s model. Third, the bank is required to have its senior management actively involved in the risk-control process. Fourth, the bank’s risk-measurement model is required to be fully integrated into the day-to-day risk-management process. Fifth, trading and exposure limits should be used in conjunction with the risk-measurement system. Sixth, the bank is required to have in place a routine program of stress testing. Seventh, the bank should have its internal policies, control procedures, and risk-management systems well documented. Finally, the bank should have an internal auditing process that regularly reviews its risk-measurement system and its risk-control unit.

Capital Requirements for Smaller Banks

Under Basle’s new approach, smaller banks that prefer not to use an internal model are required to follow a revised version of the standards proposed in April 1993. The modifications address many of the earlier criticisms on the 1993 Basle proposal. For instance, the heavily criticized vertical disallowance in the computation of interest rate risk is significantly reduced.23

The new standards cover the market risk of commodities, which is ignored in the 1993 proposal. Banks that do not use an internal model for commodity risk will have to follow either a standard approach or a simplified approach. Under the standard approach, the net position in each commodity is converted at the current spot rate into the national currency. A maturity ladder approach, similar to the one used for debt instruments, is then used to determine the capital charge. Under the simplified approach, a capital charge equal to the sum of a 15 percent charge on the net position and a 3 percent charge on the gross position of a commodity is imposed.

Box II.1.Estimates of Value-at-Risk for a Hypothetical Derivatives Portfolio

The portfolio consists of various over-the-counter and exchange-traded derivatives from three main product classes: interest rate, currency, and equity products. The products considered include interest rate swaps, Eurodollar futures and futures options, yen futures and futures options, and S&P 500 index futures and futures options. The analysis is based on actual prices and rates obtained from Bloomberg Business News on April 26, 1995. The notional principal of the portfolio is about $340 million. The replacement value of the portfolio, given the prices and rates on April 26, 1995, is $1.27 million.

For swaps, the estimation of value-at-risk involves examining extreme movements in the swap-rate curve. Two possibilities are considered: (1) parallel shifts in the swap-rate curve governed by the 90-day Eurodollar interest rate, and (2) nonparallel shifts, where movements in swap rates corresponding to different maturities are given by the differences in their sensitivities to changes in the 90-day Eurodollar rate. These sensitivities are estimated using daily data over the six-month period ending on April 26, 1995.1 Changes in the replacement value of swap positions are computed using the bootstrap method. This method infers the implied (zero-coupon) spot rates from the swap curve (that provides coupon rates for each maturity) by solving recursively for the discount rate corresponding to the last cash flow of a hypothetical par-value bond that pays a coupon rate equal to the swap rate for the maturity considered.2 Given the implied spot rates, a set of implied forward rates can be computed. The expected future payments on the floating rate side of the swap can be computed by multiplying the implied forward rates by the notional principal. The value of the swap is obtained by discounting the stream of differences between the fixed and floating payments using discount rates derived from the implied spot rates.

For futures options, three methods are considered for calculating value-at-risk: an analytic approach, which uses no approximation to the nonlinear relationship between the value of an option and the underlying asset; a delta-normal approach, which uses a first-order Taylor approximation; and a delta-gamma approach, which uses a second-order Taylor approximation. The analytic approach calculates the effect of an extreme movement in the price or rate of the underlying asset on the value of an option position by computing the theoretical difference in the option value using the exact option pricing formula. The delta-normal approach calculates the change in option value by multiplying the delta of the option by the extreme price change of the underlying asset. The delta-gamma approach is the same as the delta-normal approach except that a term (the gamma) that captures the effect of a change in the delta is added to yield a better approximation. For futures, the value-at-risk is simply computed as the change in the value of the futures position following a hypothetical extreme movement in the futures price.3

The key parameters used in the calculations are the volatilities of the various underlying assets. Various volatility estimates are used. These include historical volatility estimates based on samples of historical data of different period lengths (30 and 100 days) and implied volatility estimates that are inferred from actual call option prices on April 26, 1995.4 The various combinations of methods and parameters produce 18 different ways of estimating value-at-risk for all of the positions in the portfolio.

In the April 1995 Basle proposal, estimates of value-at-risk for different product classes are to be added together to produce the overall value-at-risk. This method of aggregation is equivalent to assuming that products from different classes are perfectly correlated. The Basle proposal allows correlations among products in the same class to be used in the computation of the value-at-risk for a particular class. Here, for each of the 18 ways of estimating individual position value-at-risk, the portfolio value-at-risk is computed using the Basle aggregation method. For comparison, estimates of value-at-risk for the individual positions also are aggregated using three alternative assumptions: (1) asset returns are perfectly correlated such that the portfolio value-at-risk is the simple sum of the individual value-at-risk of all positions; (2) asset returns are not correlated across product class, for which the portfolio value-at-risk is the square root of the sum of the squared individual value-at-risk; and (3) asset returns obey historical correlations.

Value-at-Risk for a Hypothetical Portfolio1(In thousands of U.S. dollars)
Summary Statistics2Basle Internal

Model-Based Approach
Assuming Perfect

Correlation
Assuming Zero

Correlation
Using Historical

Correlation
Mean9,7059,7968,0258,304
Maximum17,97518,13116,22816,568
Minimum3,9534,0092,5032,680
Maximum/Minimum4.54.56.56.2
Replacement value

of the portfolio3
1,270
Source: IMF staff calculations.

The portfolio is made up of the following positions: (1) a $20 million (notional) three-year interest rate swap receiving a fixed rate of 7.07 percent and paying LIBOR semiannually; (2) a $10 million (notional) four-year interest rate swap receiving a fixed rate of 7.47 percent and paying LIBOR semiannually; (3) a long position in 100 contracts of the September 1995 Eurodollar futures (the size of each contract is $1 million); (4) a long position in 150 contracts of the September 1995 Eurodollar futures call option (the settlement amount for a contract is $2,500 times the call option price) and a long position in 40 contracts of the September 1995 yen futures (the size of each contract is ¥ 12.5 million); (5) a long position in 60 contracts of the September 1995 yen futures call option with a strike price of 119 (i.e., 1.19 cents for a yen); (6) a long position in 40 contracts of the June 1995 Standard & Poor’s (S&P) 500 futures (the contract size is 500 times the S&P 500 Index); and (7) a long position in 60 contracts of the June 1995 S&P 500 futures call option (the contract value is 500 times the call option price).

For each aggregation method (column) the statistics are computed from the 18 estimates of value-at-risk.

Replacement value of the portfolio is the amount of money that would have to be paid to a third party to induce it to enter into a transaction to replace an existing deal. For a portfolio, the replacement value is equal to the present value of all profitable outstanding contracts.

Source: IMF staff calculations.

The portfolio is made up of the following positions: (1) a $20 million (notional) three-year interest rate swap receiving a fixed rate of 7.07 percent and paying LIBOR semiannually; (2) a $10 million (notional) four-year interest rate swap receiving a fixed rate of 7.47 percent and paying LIBOR semiannually; (3) a long position in 100 contracts of the September 1995 Eurodollar futures (the size of each contract is $1 million); (4) a long position in 150 contracts of the September 1995 Eurodollar futures call option (the settlement amount for a contract is $2,500 times the call option price) and a long position in 40 contracts of the September 1995 yen futures (the size of each contract is ¥ 12.5 million); (5) a long position in 60 contracts of the September 1995 yen futures call option with a strike price of 119 (i.e., 1.19 cents for a yen); (6) a long position in 40 contracts of the June 1995 Standard & Poor’s (S&P) 500 futures (the contract size is 500 times the S&P 500 Index); and (7) a long position in 60 contracts of the June 1995 S&P 500 futures call option (the contract value is 500 times the call option price).

For each aggregation method (column) the statistics are computed from the 18 estimates of value-at-risk.

Replacement value of the portfolio is the amount of money that would have to be paid to a third party to induce it to enter into a transaction to replace an existing deal. For a portfolio, the replacement value is equal to the present value of all profitable outstanding contracts.

The table above presents summary statistics of the 72 estimates of value-at-risk that result from the 18 combinations of estimation methods and parameters and the 4 aggregation methods. Under the Basle aggregation approach, the highest value-at-risk obtained is as much as 4.5 times the lowest value-at-risk estimated. For the aggregation method utilizing actual historical correlations, the highest portfolio value-at-risk estimate ($16.6 million) is 6.2 times that of the lowest portfolio value-at-risk estimate ($2.7 million).

Value-at-Risk from Alternative Aggregation Approaches(As a ratio of value-at-risk using the Basle aggregation approach)
Summary

Statistics
Assuming

Perfect

Correlation
Assuming

Zero

Correlation
Using

Historical

Correlation
Mean1.010.760.80
Maximum1.020.900.93
Minimum1.000.610.66
Source: IMF staff calculations.
Source: IMF staff calculations.

To highlight the differences across aggregation methods, The table on the left presents summary statistics of 54 ratios that, for each of the 18 combinations of estimation methods and parameters, compare estimates of value-at-risk using the last three aggregation methods (numerators) to the Basle aggregation method (the denominator). The portfolio value-at-risk utilizing historical correlations was as low as 66 percent of the value-at-risk computed under the Basle approach. On average, the ratio is 80 percent for the historical correlation method and 76 percent for a method assuming zero correlations among all assets. The portfolio value-at-risk assuming perfect correlations is close to the Basle approach because, for the portfolio considered, there is little gain from accounting for within-product-class correlations, as products in the same class are either dependent on the same underlying asset or highly correlated. This might not be the case for a more general portfolio.

Several observations can be made. First, there is variation in estimates of value-at-risk across different estimation methods and parameter inputs. Second, there is significant variation in value-at-risk estimates across aggregation methods. Third, the portfolio value-at-risk can be very large relative to the current replacement value of the portfolio. The highest value-at-risk found is more than 14 times the actual replacement value of the portfolio.

1 Essentially, a so-called one-factor model is used for the swap curve and two sets of factor sensitivities are considered in the value-at-risk analysis. The one-factor model, given its simplicity, is a popular approach for modeling the dynamics of the yield curve. Other approaches that market participants have used include a two-factor model, for which a long rate factor is added, and a two-factor model, for which a second volatility factor is included.2 A par-value bond is a bond whose current market value is the same as its principal amount.3 An extreme price movement is computed as a price movement corresponding to a rate of return that is 2.32 times the ten-day standard deviation below the mean. This corresponds to a one-tail 1 percent critical value under a normal distribution.4 The use of relatively short samples in the estimation of the underlying futures price volatility is more or less driven by the availability of data. Many short-term futures contracts are only actively traded in the two-to-three month period prior to contract expiration. For instance, on April 28, 1995, a total of 52,921 contracts of the June 1995 S&P 500 futures was traded, compared with 55 contracts of the same contract on April 29, 1994. The use of a longer sample for futures contracts, while feasible, would require contract switching, which can further complicate the estimation of futures-price volatility. The implied volatility method treats the actual option price as an input. The option pricing formula is then inverted to yield the volatility that corresponds to such an option price. When multiple options of different exercise prices are used, the implied volatility is usually found by minimizing the sum of squared discrepancies between the observed option prices and their corresponding theoretieal values.

In addition, more refined approaches are included for option risk. Specifically, gamma and vega risks are explicitly incorporated. Under the so-called delta-plus approach, a bank is required to calculate the gamma and vega for each of its option positions. For interest rate derivatives, the gammas and vegas are slotted into separate maturity ladders by currency. The capital charge for gamma risk for each time band is computed by multiplying the net negative gammas by a given risk weight and then by the square of the market value of the underlying asset. The positive gammas are ignored, because they are related to an increase in the sensitivity of the value of the derivative position to the underlying asset value when it is moving in a direction that will increase the value of the derivative asset. Hence, strictly speaking, a positive gamma does not represent risk. The capital charge for vega risk for each time band is computed by multiplying the vega by a proportional shift in volatility of ±25 percent.

Challenges for the Own-Models Approach

The new Basle proposal has several advantages. First, it eliminates the need for banks to maintain two parallel systems, one for regulatory purposes and another for trading and portfolio allocation. This will reduce costs and promote efficiency. Second, it eliminates the distortionary effects of incomplete and rigid capital rules on bank activities. Third, it promotes efficient updating of capital requirements to cover new products. Finally, it provides banks with additional incentives to improve their risk-management systems.

There are many challenges in implementing this new approach. There is very little agreement in the banking industry about how to model and measure risk; different banks employ different models, estimation procedures, and model parameters. A potential problem with the new internal-model-based approach is that two banks holding the same portfolio will most likely come up with different estimates of value-at-risk. This will imply different capital requirements for banks that incur the same risks and violates one of the major objectives of international capital requirements, namely, the achievement of a level playing field.

This problem was discovered in the second half of 1994 during an experiment conducted by the Basle Committee on Banking Supervision; the Committee asked an internationally diversified group of major banks dealing in the OTC derivative markets to estimate the value-at-risk associated with four hypothetical portfolios with and without options. Very different estimates were reported by the group of banks even though they were asked to use the same confidence interval and the same holding period. The problem with this divergence of estimates is that it can lead to issues regarding bank competitive equity and, more important, questions about model quality. For example, a bank might choose to use a model that produces a lower capital requirement rather than one that produces more accurate estimates of value-at-risk.

Supervisors in the Group of Ten countries have more or less concluded that a full-fledged, model-validation approach to bank supervision is not a practical way of dealing with these modeling issues. There is little agreement about what constitutes a good model for measuring risk, and supervisors generally lack the experience and technical knowledge required to validate the models currently in use by the more sophisticated banks. Instead, the new approach attempts to combine minimum standards, regular external auditing, a safety factor, and performance criteria in determining regulatory capital requirements for individual banks. Assessing performance might involve comparing the profit-and-loss performance of a bank to verify that its portfolio return is consistent with the ex ante value-at-risk estimates produced by the bank’s model. If a bank experiences losses in excess of its ex ante estimates of value-at-risk more often than what is implied by the underlying confidence interval used in estimating value-at-risk, then the bank’s internal model might fall short of acceptable quality. In such cases, capital requirements greater than “value-at-risk times three” would be imposed. This would be achieved by setting the safety factor above the minimum of three.

A strength of this performance-based approach appears to be that the need to examine and validate a bank’s model would, to a certain extent, decrease. A weakness of this approach is that ex post performance analysis is usually not very powerful statistically because it requires a long history of stable market behavior to judge the performance of a model. Moreover, past performance might not reflect future model performance because banks continuously change their models to accommodate changes in the market environment, the introduction of new products, and the introduction of new pricing and risk management techniques.

Although Basle’s draft proposal suggests that the size of a bank’s safety factor will be related directly to the ex post performance of the bank’s model, it does not provide details about how the safety factor is to be determined. Setting the right criteria for assessing model performance is key to providing banks with incentives to improve model accuracy. In short, the effectiveness of Basle’s model-based approach depends crucially on whether this incentive structure encourages banks to choose model accuracy over lower capital requirements. It is likely that the Basle Committee will have to establish a framework in which capital requirements can be related directly to performance before the new model-based approach can be fully implemented.

An alternative approach to performance evaluation, the so-called precommitment approach, requires that each bank commits to a ceiling level of capital losses that will not be exceeded.24 Should this ceiling be breached, a severe penalty would be imposed on the bank. This alternative would provide the incentive for banks to dynamically control their risk exposures by accounting for changes in future portfolios and strategies. Model validation is not necessary under this alternative approach.

A potential difficulty in implementing the pre-commitment approach is that it may be as difficult to establish rules for assessing penalties—a penalty function—as it has been to establish simple capital adequacy standards. For example, in principle, the penalties should reflect the cost to taxpayers of a bank failure or the indirect costs of an increase in future funding costs should a bank violate its capital precommitment or both; each is difficult to measure. Furthermore, given that the market environment is changing all the time, it also would be difficult to establish a robust penalty function. Another potential problem involves credibility. A bank would violate its capital commitment only because it has suffered a big loss. In such cases, regulators would be unable to credibly impose a sizable penalty on a bank that is at risk of failing. In fact, a huge penalty might force a bank to adopt a go-for-broke strategy after losing a significant amount of its capital. Flexibility in applying penalties is another issue. A bank might violate its capital commitment because of bad luck or because of a systemic event. In such cases it would not be appropriate to penalize an otherwise safe and sound bank. But if application of the penalty function is flexible, then a measure of the quality of the bank’s risk-management model is necessary and supervisors will then have to validate the bank’s model.

There are also open questions about implementation of the new Basle approach. The emphasis on quantitative measures, like value-at-risk, might lead banks to miss major market events that are not easily quantifiable. A recent example of such an event is the Mexican financial crisis. Many market participants and regulators have serious doubts about whether a statistical-mathematical model could have detected the increase in risk associated with this crisis. While Basle’s new approach would require stress testing, there is no consensus about how the results of stress tests should be incorporated into the estimation and analysis of value-at-risk.

Another issue is that the new Basle approach does not deal with legal risk and operational risk. The collapse of Barings and the legal problems encountered by Bankers Trust have revealed that such risks can be important. Furthermore, even though the Basle approach incorporates liquidity risk—a ten-day holding period in the computation of the value-at-risk—a uniform treatment might not be appropriate because changes in liquidity depend on the instrument in question, the size of the position, and the market conditions. By treating each position identically, bank portfolio decisions might well be distorted; the risk of holding a very large concentrated position, which is often difficult to liquidate, might be underestimated.

International and Bank-Nonbank Coordination Issues

The Basle Committee is planning to adopt the new approach by the end of the year and to implement it by the end of 1997.25 Meanwhile, the CAD, which does not allow the use of internal models for determining capital requirements, is required to be implemented by EU countries by January 1, 1996. The EU has tried to accommodate the new internal model-based approach proposed by the Basle Committee by agreeing to allow top European banks to follow the Basle approach to some extent. In April 1995, the commission’s directorate-general has agreed that daily value-at-risk models can be used in conjunction with the CAD provided that the capital requirement according to the firm’s estimate of value-at-risk is not less than the capital requirement according to the CAD. In any event, should the Basle proposal be adopted, the EU has to amend its CAD sometime after its implementation. Without careful harmonization, some EU banks might have to adjust their risk-management system twice; once to meet the CAD deadline, and once to adopt Basle’s new approach in 1997.

Also at issue is whether there will be a level playing field in areas where both banks and securities houses compete. This issue is more serious for U.S. banks than for European banks, which are universal banks. It is unclear whether, and to what extent, securities house regulators will allow securities houses to use internal models to determine capital requirements.

The Basle Committee and the International Organization of Securities Commissions (IOSCO) have tried to coordinate efforts on capital requirement issues but have not had much success. A major difficulty is the difference in the regulatory focus of bank and securities regulators. Bank regulators tend to focus on systemic risk issues as banks are relatively more vulnerable to contagious collapses; securities regulators tend to pay more attention to investor protection issues. As a result, securities regulators concentrate primarily on liquid capital that measures the ability of a securities firm to meet its obligations to investors and creditors.

The Focus on Process of Internal Risk Management

Initiatives to Improve Risk Management

Regulators have come to recognize that maintaining a minimum capital requirement is just one part of the overall risk-management system of a bank. In a report on derivatives released in May 1994, the U.S. General Accounting Office recommended that regulators should assess the quality of the major OTC derivatives dealers’ risk-management systems.26 Some market participants have argued that having a top-rated, risk-measurement model but a poor risk-management and -control process could be worse than having a poor internal model but a good risk-control process.27

A recent survey by the Group of Thirty has shown that, in general, dealers do not follow uniformly the recommendations on sound risk-management practice of the Group of Thirty report issued in July 1993.28 The earlier report contained recommendations in five major areas: (1) general policies; (2) valuation and market risk management; (3) credit risk measurement and management; (4) enforceability, systems, operations, and controls; and (5) accounting and disclosure.

The survey examines to what extent the recommendations in the July 1993 Report are followed by market participants. On the measurement and management of market risk, only 21 percent of dealers measure and compare market-risk exposure against established limits on an intraday basis. Regarding the use of value-at-risk for measuring market risk, only 43 percent of the dealers surveyed employed this approach, and fewer than half of those who employed the value-at-risk approach used a consistent confidence interval and time horizon for different derivative portfolios. Furthermore, only 54 percent of the dealers performed stress tests and simulations to measure how their portfolios would perform under abnormal changes in market conditions. In addition, in conducting stress tests, only 27 percent of the dealers account for decreases in liquidity and only 28 percent account for failures of a significant counterparty.

On the operation of an independent market-risk function, the survey revealed that only 29 percent of the dealers monitor the variance between a portfolio’s actual and predicted volatility. Furthermore, more than 20 percent of the dealers do not establish counterparty credit limits by current and potential exposure.

To further promote the importance of having a high-quality risk-management system and to establish a benchmark for good risk management, the Basle Committee on Banking Supervision and IOSCO issued under the same cover, in July 1994, two separate reports on risk management regarding derivatives.29 Although styles differ significantly, due to traditional differences in the supervision of banks and nonbanks, there are only minor differences in the substance of the recommendations.

The reports identified many important elements of sound risk management. First, senior management must be involved in setting the overall policy of the firm regarding the use of derivatives including the level of risk tolerance and the range of products allowed. Second, a line of authority and responsibility for managing risk and conducting derivative activities should be clearly established. Third, before a new derivative activity is undertaken, a detailed proposal describing the product, market, and strategy should be prepared. A bank should not get involved significantly in a new product until it has been fully integrated into the bank’s risk-measurement system. Fourth, compensation policies should be set up in a way to avoid excessive risk taking. Fifth, banks should have their internal auditors to evaluate compliance with risk limits. If limits are exceeded, the system should automatically notify senior management.

Sixth, the internal auditors should check for adequate separation of duties and oversight. Seventh, the reliability and timeliness of information reported to the bank’s senior management and the board of directors should be emphasized. In particular, the measured risk for derivative activities should be translated from a technical and quantitative format to one that can be easily understood by senior management who might not have the technical background. Eighth, risk should be measured and aggregated on an institution-wide basis including both trading and nontrading activities. That is, it is not sufficient to concentrate on the trading book. Ninth, stress tests should be conducted. Stress analysis should include the identification of possible adverse events, an evaluation of the likelihood of adverse events, the analysis of worst-case scenarios, and an evaluation of the bank’s ability to withstand the worst events. Stress tests should also include qualitative analysis of hard-to-quantify events and the actions management can take under those scenarios. Tenth, the institution should regularly review the underlying methodologies and assumptions of its models.

Difficulties in Supervising Internal Risk-Management Systems

Supervisors face many challenges in examining the internal risk-management systems of banks. First, supervising a bank’s risk-management system is significantly more complicated than determining a bank’s regulatory capital requirements. Second, the quality of different aspects of a risk-management system is difficult to quantify and to aggregate to obtain an overall measure of the quality of the risk-management system of a bank. For instance, while it might be easy to tell whether a bank has performed stress testing or not, it is very difficult to judge the quality of the stress-testing exercise. Even banks might not agree among themselves on what is a reasonable set of stress scenarios. Assumptions on the holding period and liquidity can also significantly affect the stress test results. In addition, stress testing of derivatives-like options is difficult, because the pricing of options is generally based on certain distributional assumptions. Large jumps in prices might not be consistent with these assumptions. Third, there might not be one optimal risk-management system for banks of all size and business emphasis. The system that is optimal for a large bank that does significant trading in the OTC derivative markets is probably not optimal for a smaller bank that concentrates on more traditional banking business. It would be difficult for the regulators to design specialized risk-management benchmarks for different banks.

II Initiatives Relating to Derivatives

The sudden collapse of Barings has heightened concerns about whether existing transparency and disclosure requirements for the derivative activities of individual financial institutions are adequate for judging the safety and soundness of individual financial institutions and for safeguarding financial systems against systemic problems. It is possible that had there been greater information about the derivative positions taken by Barings, the investment bank might have been prevented from accumulating positions risky enough to put at risk the entire capital of the firm.30 The lack of transparency also can lead, during periods of market distress, to overreactions by market participants and prevent policymakers from making timely decisions about how to contain the effects of extreme events when they occur.

The misfortunes of various end-users also have raised concerns about the adequacy of investor protection in the unregulated OTC derivatives markets.31 In several cases, end-users have accused dealers of selling high-risk derivative instruments that were not suitable for their purposes and for withholding information about the riskiness of the derivative instruments. Many end-users, especially those lacking expertise, also are concerned about their current exposures, which they fear to be unsuitable for their purposes and the result of aggressive marketing by dealers.

Motivation for Industry and Regulatory Initiatives

Derivative markets have grown rapidly in recent years. In addition to this rapid growth, many sectors of the economy have made extensive use of derivative instruments; even firms engaged in nonfinancial activities are now using derivatives, especially OTC derivatives. Derivative instruments obtain their value from the value of the underlying securities, interest rates, or price indices, and the reasons for their popularity are clear. If used properly, derivative instruments are useful and convenient for hedging against many kinds of risk, including fluctuations in interest rates, exchange rates, and commodity prices, for example oil prices, and for unbundling risk so that risks can be distributed to those willing to incur them. OTC derivative contracts are especially popular because they are privately negotiated contracts between two parties and often are custom designed to suit the specific needs of counterparties. Thus, OTC derivative contracts can be used to achieve highly specialized funding and risk-bundling objectives.

For a number of important reasons, a financial institution’s active participation in OTC derivative markets might reduce the transparency of the institution’s financial condition and risk exposures. First, OTC derivative positions are held as off-balance-sheet items. As a result, very limited information is available about a firm’s derivative activities, making it difficult to judge what part of a firm’s profitability and risk exposure is attributable to these activities. Second, most accounting standards were established before the introduction of many OTC derivative instruments and were not designed to account for the risk exposures and income flows associated with these instruments. Third, the risk exposure of a derivative position cannot be assessed without knowing all related positions and the exact nature of the dynamic trading strategy employed. Fourth, the value of derivatives can be sensitive to changes in the prices of the underlying securities, and the value of an institution’s derivative portfolio can change dramatically and rapidly.32 The speed with which a firm’s financial condition and risk exposure can change has significantly reduced the value of annual or semiannual reporting of income statements and balance sheets. Finally, OTC derivatives are privately negotiated, custom-designed products, which may involve many counterparties in different countries; these features make it very difficult to gather and maintain market-wide information.

It is generally recognized by regulators and market participants that without transparency it is difficult to assess the safety and soundness of financial institutions and, therefore, to limit systemic risk. Although regulatory capital requirements and internal risk management are necessary for safeguarding the soundness of financial institutions, they are by themselves insufficient for eliminating systemic risk. For example, a sound financial condition alone cannot prevent a massive withdrawal of funds from an institution at a time of market distress unless depositors and creditors are fully informed. As in the case of a bank run, a massive withdrawal of credits can render an otherwise solvent and sound bank insolvent if it has to liquidate its assets quickly, at highly discounted values, to satisfy its need for liquidity.

There are other advantages to greater transparency. Institutions known to have a good risk-management system might face lower funding costs (including higher stock prices), receive greater credit lines, and attract more business. Thus, transparency can provide incentives for financial institutions to improve their risk-management systems. Greater disclosure also would allow market participants to obtain better information and to better assess the risk exposures of its counterparties. Good information is essential for making trading and risk-management decisions, such as counterparty position limits, and also allows investors to make better investment decisions. Thus, greater transparency and disclosure can improve the allocative efficiency of financial markets.

Not all market participants agree, however, that more transparency is beneficial. An important impediment to greater transparency and disclosure has been the reluctance on the part of some market participants to disclose too much information about their financial positions, risk exposures, pricing models, and risk-management systems, This proprietary information is viewed as the source of revenue and market shares and, therefore, should not be disclosed. This is particularly true for large, sophisticated, and successful market participants who utilize leading edge pricing and risk-management technologies and who have large exposures in a wide variety of markets. Another impediment to greater transparency and disclosure is that there is no consensus about what information to disclose, when to disclose it, and to whom to disclose it.

Initiatives to Deal with Transparency

Initiatives have been taken in the past year by some of the major industry and regulatory bodies in the financial industry to improve transparency and disclosure. These initiatives included the Institute of International Finance (IIF) proposal, issued in August 1994, for a framework for public disclosure of derivatives activities and related credit exposures;33 the Bank for International Settlements (BIS) report, issued in September 1994, on the public disclosure of market and credit risk by financial intermediaries (the Fisher report);34 and the U.S. Financial Accounting Standards Board (FASB) statement (Statement of Financial Accounting Standards No. 119), published in October 1994, on the disclosure of derivative positions. After reviewing these initiatives, many of the unresolved issues involving the disclosure of information by individual institutions are examined.

Institute of International Finance Report

The IIF proposal focuses on the disclosure of credit risk in annual reports and develops an international standard for public disclosure of derivatives-related credit exposures. The report emphasizes the importance of disaggregated data on credit risk exposures and activity levels, and specifically recommends that replacement value (or net replacement value if netting arrangements are enforceable) be disclosed by categories of counterparty credit quality based on credit ratings by rating agencies, internal credit ratings, or the 1988 Basle Accord classification based on the OECD membership.35 For example, replacement value would be reported according to the private credit ratings AAA, AA, A, B, and non-investment grade. When internal credit ratings are used, information would be required to be reported on the institution’s credit-scoring system.

The report also calls on financial institutions to provide more detailed information about their derivative activities, including notional values by product types—such as interest rate contracts, foreign exchange contracts, and equity contracts—and their maturities. Under the category of interest rate contracts, for example, an institution should report its exposures in swaps, options, futures, forwards, and other derivative products, and according to whether they are OTC or exchange-traded products. The total market value for each product type—the sum of positive mark-to-market values—also would be disclosed. The report does not recommend adjustments for netting and collateral arrangements in the calculation of market value, however, because they are difficult to quantify.

Some market participants have criticized the approach recommended by IIF because it uses notional value rather than replacement value as a measure of market size. The relationship between these measures can vary significantly from institution to institution. For example, at the end of 1993, the ratio of gross replacement value to notional value was 3.4 percent for one firm and 0.5 percent for another.36 In addition, the notional value of interest rate swaps and currency swaps cannot be compared because the notional principals are exchanged at maturity for currency swaps but not for interest rate swaps. The report has been criticized for other reasons as well: the usefulness of annual reporting is limited for volatile derivative contract, and credit-risk exposures cannot be accurately represented by the total current replacement values for all derivative contracts because they ignore potential future exposure. OTC contracts are marked-to-model, which cannot be verified by outsiders; most OTC derivative products are custom designed and cannot be traded in liquid markets, which makes it difficult to value them.

The Fisher Report

The Fisher report addresses the disclosure of both market and credit risk and emphasizes the reporting of market risk and the performance of risk-management systems. Its recommendations on credit risk are in line with those of the IIF report. The Fisher report also recommends the reporting of potential future credit exposures, but it does not provide guidelines on how to compute and report potential credit exposure.

Among the important features of the report is its endorsement of value-at-risk as a measure of market-risk exposure. It makes specific recommendations about how value-at-risk should be disclosed to reveal the risk appetite and risk exposure of an institution, such as providing high, low, and average measures of value-at-risk for different holding periods (say one day and two weeks) during the reporting period; confidence interval at which value-at-risk is computed; a comparison of the average value-at-risk for a holding period of one day and the average change in portfolio value; and measures of the volatility of the change in portfolio value. The report also recommends comparisons of ex ante value-at-risk estimates to actual losses incurred by the institution to observe the frequency with which actual performance is worse than predicted by the firm’s risk-management system.

A key issue not addressed by the report is the comparability across financial institutions of their estimates of value-at-risk. As discussed earlier, there are many ways to compute value-at-risk and estimates depend on the holding period, the confidence interval, the derivatives-pricing models, and the parameter inputs chosen. As such, two firms with the same portfolio might produce very different estimates of value-at-risk. Another issue is that value-at-risk is a snapshot of a firm’s risk exposure and cannot reflect the effects of any dynamic hedging strategies that the institution might be using.

Statement of Financial Accounting Standards

In October 1994, a new accounting rule for derivatives, SFAS No. 119, was issued by the U.S. Financial Accounting Standards Board.37 Inappropriate accounting rules for derivatives often have been cited as a major impediment to better disclosure in OTC derivative markets. Previous rules related to derivatives were deemed insufficient for many complex OTC derivative instruments.38 A problem with traditional accounting standards is that they were designed using an instrument-specific approach that is not appropriate for OTC derivatives, which often cut across different product classes. In addition, many accounting standards were drafted before the introduction of many OTC derivative products. Moreover, many standards require differential treatment of derivative positions for hedging purposes and for other purposes. Under the so-called hedge accounting approach, positions for hedging are reported together with the position being hedged and no marking-to-market is required. But a derivative position can also be used to partially hedge a portfolio, which introduces ambiguity about whether a position should be identified as a hedge for accounting purposes.

The new accounting standard, SFAS 119, applies to all U.S. business enterprises and nonprofit organizations and covers derivative instruments including futures, forwards, swaps, and options and similar contracts. It does not include on-balance-sheet instruments like structured notes. The standard applies differential treatment for derivatives used for trading and derivatives used for other purposes.

For derivative positions for trading purposes, SFAS 119 requires (1) disclosure of the average fair value of the position during the reporting period and at the end of the period, (2) distinction between positive and negative values to prevent institutions from netting winning and losing positions unless they are with the same counterparty, and (3) disclosure of net gains and losses disaggregated by product types and risk classes. For other derivatives, the standard requires (1) disclosure of the objective for holding the positions, (2) an explanation of the strategies used to achieve the objective, and (3) a description of the reporting methodology. For those derivatives receiving hedge-accounting treatment, the end-user must also provide a description of the underlying positions being hedged and the amount of deferred gains and losses.

In contrast with the recommendations of the IIF and the Fisher reports, SFAS 119 focuses on the disclosure of qualitative information. In particular, it does not require disclosure of value-at-risk, stress analysis, durations, and details on current positions. The FASB took the position that a quantitative measure of risk should not be required unless there is general agreement on the most appropriate measure of risk and the best way to calculate it. Many market participants feel this approach is conservative, however, and think that imperfect quantitative disclosure is better than none at all.

Challenges for Improving Disclosure

The above-mentioned reports have made significant progress in raising important issues involving the disclosure of credit and market risk. There are still a significant number of challenges, of a more general nature, that the industry and regulators are likely to face in the period ahead. All of these challenges point out the difficulties involved in defining a single measure of risk that captures comprehensively all of the risks inherent in the financial activities of a single financial institution. These difficulties pose major challenges for internal risk management, for supervisors in assessing safety and soundness, and for assessing the potential for systemic risk.

Risk Aggregation

Quantitative disclosure requirements have focused mainly on the trading book of an institution. A case has been made by some market participants for not treating the trading book separately, however. One reason is that a more comprehensive picture of the risks that a financial institution is incurring is necessary to properly assess the safety and soundness of the institution. An additional problem is that because credit and market risk are not independent, they cannot be measured separately and then simply added together. In general, how to aggregate market and credit risk in order to obtain a comprehensive picture of a firm’s risk exposure is a question that still needs to be answered.

Nonquantifiable Risks

In addition to market and credit risk, there are liquidity, legal, and operational risks; all of them can be significant. The failure of Barings can be attributed to operational risk, the risk of losses due to mismanagement, human error, and lack of control. Legal risk, uncertainty about the validity and enforceability of contracts, has also become important. For example, there is considerable uncertainty about whether some netting arrangements are enforceable, especially when they involve institutions from more than one country, and there is also uncertainty about the validity of the transaction itself. After suffering major losses from derivatives trading, various end-users have sued or accused dealers of unauthorized trading, misrepresentations of risks inherent in instruments, and not disclosing valuable information on pricing and risk.39 Liquidity risk, uncertainty about whether a position can be liquidated rapidly at current prices, has also increased in importance. Although there are indicators of market liquidity, such as bid-ask spreads and market-trading volume, they tend not to be accurate measures of risk in times of market distress. Moreover, it is difficult to estimate how quickly liquidity can change under extreme market conditions.

Because of the dearth of good measures of these kinds of risk, it is difficult for financial institutions to provide senior management, shareholders, and counterparties with accurate assessments of the riskiness of their portfolios. In addition, these nonquantifiable risks are not independent of market and credit risk. When there have been legal disputes, the party has sued after a significant increase in market risk or market event. Moreover, an increase in credit risk may be associated with an increase in liquidity risk. When the market has information (or hears a rumor) that a major counterparty might default on its obligations, the counterparty may lose its access to existing credit lines and credit markets in general.

Quantitative Versus Qualitative Disclosure

There is some disagreement about the relative importance of quantitative and qualitative disclosure. Numerical estimates of risk exposure are crucial for judging the financial condition of a financial institution. Some market participants caution, however, that a single number might be misleading especially since there are important measurement and aggregation problems. In addition, focusing on one aspect of risk that is quantifiable might detract attention from the other, and in some cases more important, less-quantifiable risks. Qualitative information, such as detailed consultations with management and analyses of accounting, risk management, netting, trading, limit setting, collateral, and how derivatives activities fit into the firm’s business, can be more useful than reporting a single measure like, for example, value-at-risk.40

Uniformity and Flexibility

There is a trade-off between uniformity (or comparability) and flexibility in disclosing information. Incomparability makes it difficult for market participants to distinguish sound from less sound institutions. In times of market distress, this inability to accurately distinguish the riskiness of one institution from another can lead to a massive and uniform withdrawal from all institutions at the same time, and create systemic problems. Incomparability might also lead to higher funding costs—through risk premiums—for all institutions, thus reducing allocative efficiency. Incomparable disclosure would also limit the government’s ability to implement appropriate policies.

Market participants might not always face the kinds of incentive structures that encourage them to disclose comparable information. For example, according to market analysts, institutions may deliberately confuse their competition by disclosing noncomparable data. This might be why one bank discloses gross replacement value as a measure of credit risk exposure, another discloses net replacement value, and still another discloses net replacement value net of collateral for credit risk disclosure.

There is also disagreement about whether rigid standards are appropriate. Some market participants have noted that rigid standards cannot accommodate changes in market structure, such as the introduction of new products, and there is still considerable disagreement about the best method for measuring risk. There is also the concern that the development of a comparable risk measure might imply the choice of an industry standard that is not necessarily the most accurate risk measure. For example, it is generally agreed that net replacement value is more accurate than gross replacement value as a measure of credit risk; however, the net measure is less comparable than the gross measure because some institutions might have more netting agreements and there might be differences in legal judgment on the enforceability of some netting agreements. In addition, net replacement value cannot be easily broken down by maturity, product type, and currency. Finally, different countries have different accounting and reporting standards. It might not be fair or feasible to impose the same standard on market participants from different countries.

Disclosure of Risk-Management System

Many market participants and regulators have found it useful to have information on the quality of the risk-management system of an institution, as risk exposures can change significantly and rapidly. The Fisher report recommends using a comparison between ex ante risk estimates and ex post losses as a way to test the performance of a firm’s risk-management system. This approach requires a long history of past performance in similar market environments and might not produce a good indicator of future performance in different market environments. Moreover, this approach might not provide very precise information. Given that institutions are likely to use different risk-management systems and different derivative-pricing models, an important challenge is to determine what to disclose about risk management. An equally important and difficult challenge is to design a common framework for the reporting of risk-management systems and processes.

How Much Disclosure Is Enough?

An important issue in disclosing information about risk-management systems is where to draw the line between adequate and excessive reporting. In the dispute between Bankers Trust and Proctor & Gamble, the latter had accused the former of not sharing its complex pricing model used in pricing the derivatives it had transacted. This raises the question, Should a dealer be required to disclose proprietary information, including its pricing model, to its customers? From the end-users’ point of view there are no market prices for OTC derivative products and they should have access to information about how the dealer values these instruments. From the dealers’ point of view, the proprietary model is a trade secret. In another dispute between Bankers Trust and Gibson Greetings, the latter accused the former of not disclosing information about the effects of volatility on the market values of its positions. This raises another question: How many stress tests are enough? Given that there are many ways to design stress tests and many scenarios, it would be difficult and costly for dealers to examine every possible case. On the other hand, end-users would want to know whether the dealer has selectively omitted stress tests in their reporting and misrepresented the risks involved in the transaction.

A related question is whether more disclosure is beneficial. Many end-users and investors have found that a large amount of information can be overwhelming and distract them from focusing on key issues. In addition, there are obvious economies of scale on information collection and maintenance. Large institutions, active in OTC derivative markets, most likely have very sophisticated reporting systems in place. The cost of extensive reporting to these institutions is relatively low. However, the cost of the same level of reporting can be very high for a relatively small institution. Requiring the same amount of disclosure for all institutions might tilt the playing field against the less sophisticated players. What the optimal amount of information to disclose is and whether differential standards should be applied to different institutions are important questions to be answered.

Frequency of Reporting

Another important issue is the frequency of reporting. The consensus of market participants and the regulators is that low frequency reporting, like annual or semiannual reporting, is not as useful as in the past. The question then is, What is the optimal frequency of reporting? The answer is not independent of the amount of information to be disclosed or reported. The optimal trade-off between the two is an unresolved issue.

Who Should Report and to Whom?

Although many of the large, sophisticated banks obtain a large share of their revenues from derivative activities, many banks, large and small do not. A key question is, Should all banks be subject to the same reporting requirements for their derivative activity regardless of the scale of their activities? One reason why it might be beneficial to have differential requirements is the cost involved in maintaining accounting and auditing for these systems, and there may be economies of scale that cannot be obtained in these areas by banks not actively engaged in derivative markets. On the other hand, from the perspective of providing investor protection, it is difficult to justify this differential treatment for disclosure requirements.

There are three potential groups that would benefit from greater transparency and disclosure: regulators, counterparties, and the general public, including creditors, investors, and shareholders. A problem is that each of these groups might find useful different types of information. If investor protection is the key concern, then disclosure of derivative activities and risk exposures to counterparties and shareholders would be justified. To control systemic risk and to contain market reactions during periods of market distress, disclosure of all this information to the general public would be useful. However, there is a trade-off between disclosing a lot of sophisticated information that overloads the average investor and not supplying enough to the sophisticated investor or counterparty. In addition, it is not agreed what the best channel is for distributing a large amount of information to the general public. While electronic networks can be very useful, most investors still do not have access to such facilities.

Currently, the U.S. Securities and Exchange Commission (SEC) is considering a two-tier reporting system. The idea is that more frequent and detailed information is reported first to the regulators on a confidential basis, and the regulators in turn then aggregate and simplify this information and provide it to the general public. What is undecided is just how regulators can transform this core data into an information base that is useful for the many different kinds of investors. Also undecided are how often regulators should receive this information and how the information ought to be provided to the general public.

Role of Credit-Rating Agencies

Credit ratings by private agencies are viewed by some market participants as providing the same benefits to the general public that greater information disclosure would. Given that funding costs are often dependent on credit ratings, financial institutions have sufficient incentives to provide information to the rating agencies, and to appropriately modify disclosure to the agencies to reflect changes in their business activities. Some market participants have suggested that private ratings are a more flexible and incentive-compatible way of satisfying disclosure requirements. One potential problem with this approach is that a single credit-rating agency might not require all of the information necessary to accurately measure the complex risk exposures of a firm. Furthermore, the rating itself might not reflect the financial condition of an institution during a period of market distress; for example, Orange County, California, went from a rating of AA to a rating of junk in one day.

There are also important issues that arise from the growing dependence of financial regulations on credit ratings. The 1993 Basle capital proposal, for example, imposes lower capital charges on instruments with investment grade ratings, and the U.S. SEC’s net capital rule gives preferential treatment to some instruments with investment grade ratings by at least two nationally recognized rating agencies. The important role now played by rating agencies may be such that regulators would find it necessary to evaluate their performance and approve their activities. But, if regulators were able to “rate” the rating agencies, then why would they need to rely on the rating agencies in the first place?

Macro Disclosure

As important as institution-specific information can be for assessing the safety and soundness of an institution, the availability of market-wide information on derivatives—what is meant by “macro-disclosure”—is equally important for assessing the condition of derivative markets. Currently, there are no comprehensive data for these markets. How then can policymakers assess the extent to which a market disturbance would have the potential to create a systemic problem in the derivative markets and in other related markets? Equally important is the ability of regulators and market participants to properly assess risk and act on these assessments during a crisis situation.

An important impediment to collecting market-wide information on derivatives is that data across a broad range of financial institutions are not comparable. Noncomparability also makes it difficult to aggregate data across institutions, and across institutions from different countries, where there may also be differences in accounting practices and inconsistencies in report formats. In addition, differences in reporting cycles create aggregation problems and problems of interpretation of the resulting data.

The BIS released a report, in February 1995, on how to improve macro-disclosure (The Brockmeijer report).41 The report emphasizes the importance of coordination by central banks in the collection of timely, global, aggregate data for derivative markets. The report called for an extension to derivative markets of the Triennial Survey of Foreign Exchange Markets, conducted by central banks and coordinated by the Bank for International Settlements. Data on notional size, geographical distribution, gross replacement value, and turnover would be collected across a wide range of countries. The report also recommended that the largest financial institutions in the derivatives markets, both banks and nonbanks, should be surveyed regularly on a consolidated basis. The information to be collected should include the concentration of activity and exposures, and market liquidity, which can be collected from turnover data and real-time monitoring of average transactions size and bid-offer spreads. The BIS launched in April 1995 a statistical survey on activity in the derivatives markets covering 26 countries as a part of the Triennial Foreign Exchange Survey.

Investor Protection and Suitability Issues

The rapid growth in the popularity of OTC derivative contracts has led to concerns about investor protection. Complex derivative products are created and priced by in-house technical experts at the major dealers, using sophisticated mathematical models, proprietary data, and high-powered computers. Were it not for these dealers, who reap substantial economies of scale, most end-users would be unable to afford to create these instruments. Because of the custom-design nature of many OTC derivative products, there generally is not a readily available secondary market for many of the OTC products. The dependence of end-users on dealers for both product design and for “market-making” makes them vulnerable to legal and operational risk and to fraud. In addition, there is the problem of adverse incentives; dealers earn higher fees by selling more complex and leveraged products to end-users. Moreover, the compensation of traders is related to the revenues they generate rather than to the risks that they take.

In part in reaction to these concerns, the Federal Reserve Bank of New York released a code of conduct, in January 1995, titled “Wholesale Transaction Code of Conduct.”42 The code proposes, but does not require, general rules of conduct for the relationship between derivative dealers and end-users. Participants are encouraged to evaluate counterparty capability, either internally or through independent professional advice, in order to understand and make independent decisions about the terms of its transactions. The code recommends that should a dealer determine that a counterparty does not have the capability to understand certain transactions, it should either enter into a written agreement stating that the end-user is relying on the advice of the dealer or simply not enter into the transaction. In addition, if a dealer decides that its counterparty does understand the transaction, but the nature or the riskiness of the proposed transaction seems to be inappropriate for the counterparty, it should then inform the counterparty of its views and document, in writing, its own analysis.

Some dealers have expressed concern that the voluntary code of conduct might ultimately become an enforceable code, which would imply higher operating costs. In addition, it is feared that adherence to the code, which can be judged on a subjective basis, might be used as court evidence when there is a dispute between a dealer and an end-user. The code is currently under review by the members of the International Swaps and Derivatives Association, Inc. (ISDA), the New York Clearinghouse Association, the Public Securities Association, the Securities Industry Association, the Foreign Exchange Committee of the Federal Reserve Bank of New York, and the Emerging Markets Traders Association.

In addition to this code of conduct, two other codes of conduct were released in March 1995. One is by a Derivatives Products Group (DPG) formed by six major securities firms—Goldman Sachs, Salomon Brothers, Morgan Stanley, CS First Boston, Merrill Lynch, and Lehman Brothers. The other code was issued by the ISDA. The DPG has agreed with the U.S. SEC and the U.S. Commodity and Futures Trading Commission to disclose their derivative exposures and marketing approaches for derivatives. The DPG has also agreed to provide end-users written statements about the risk of derivatives. The ISDA code also emphasizes relationships with customers and the mechanics of derivative transactions.

III Mechanisms for International Cooperation in Regulation

The increased international integration of financial markets and recent advances in communications technology mean that a financial crisis is unlikely to remain isolated to one institution and one national jurisdiction. The prevention and the containment of systemic problems now require the cooperation and coordination of supervisory and regulatory authorities across national boundaries.

Multilateral Banking Supervisory Organizations

Basle Committee on Banking Supervision

Among the several banking supervisory organizations, the Basle Committee on Banking Supervision is the most recognized.43 The Basle Committee was established in 1974 by the central bank governors of the Group of Ten and now consists of representatives from the bank supervisory authorities and central banks of 12 countries.44 The Bank of International Settlements facilitates meetings and provides secretariat support for the Committee.

The Basle Committee often establishes working groups to examine specific issues. Frequently the role of the working groups, headed by a chairman, is to produce discussion papers or policy reports for review by the Committee. Consensus decisions regarding the content of the reports are made within the working group, with the chairman providing clear direction. The Committee releases the report for comment, discusses the submitted comments, and, if it deems appropriate, reissues its recommendation as a guide to best supervisory practice. The Committee has no enforcement authority and implementation takes place through members’ national regulatory structures. To assure widespread adoption of its recommendations, the Committee strives for consensus.

European Union

The European Union (EU) formally entered the banking supervisory arena in 1977 with the enactment of the “First Banking Directive,” setting the stage for regulatory coordination among EU members.45 Of primary importance in the area of banking supervision are the European Commission (the Commission), the body that initiates legislative actions, the Council of Ministers (the Council), the body that ultimately approves or rejects proposals for directives, and the European Parliament, a consultative body in the area of bank supervision.46

The Banking Advisory Committee (BAC), established in the First Banking Directive, advises and assists the Commission in the area of banking.47 An informal contact group of the EU Banking Supervisory Authorities (the Contact Group) meets and discusses a wide range of topics related to national supervisory developments. Descriptive papers on the discussed topics are presented to the Banking Advisory Committee. Banking supervisory issues are also discussed within the European Monetary Institute (EMI), the future central bank for the EU. The EMI has no executive responsibility for prudential supervision but is expected to examine issues of a macro-prudential nature and to be consulted on legislation that influences financial market stability. The Banking Supervisory Sub-Committee assists the Council of the EMI.

Unlike the recommendations of the Basle Committee, EU Directives are binding and oblige the member states to adapt their national law to the directives. The process by which a directive becomes EU law is as follows.48 All legislation for the EU must be initiated by the Commission. The relevant committees, staff of the Commission, and other experts aid the Commission department in drafting proposals. The Commission votes on the draft proposal: passage occurs with a simple majority.49 The Commission’s proposal is then sent to the Council. The Council’s “Economic and Finance” (ECOFIN) formation, composed of finance ministers, considers banking proposals, following the new “co-decision” procedure. First, the Council requests opinions from the Parliament and the Economic and Social Committee, an ancillary body that must be consulted on economic matters. Using their opinions, the Council then adopts a common position and transmits it to Parliament for a second reading. If Parliament approves the common position without amendment, it may be approved by the Council by a qualified majority50 without being altered by the Commission. However, if Parliament rejects the common position or wishes to make amendments, the new co-decision procedure provides for a “Conciliation Committee,” which attempts to negotiate a compromise between the Council and Parliament. If agreement is still not attained, Parliament has veto power; its rejection means the proposal is lost.

Compared with the consensus decision process of the Basle Committee, which consists of a working group report and the Committee’s final recommendation, the EU Directive process requires a more structured interactive process of several bodies, including the Commission, the Council, the Parliament and their attendant committees. Because the EU enacts binding legislation, the checks and balances among its different institutions purposely provide for broad participation by the member states that will be required to enact national legislation.

The Basle Committee and the EU Directive processes are compared in Figure 1 for the issue of monitoring and controlling large exposures of credit institutions, a subject on which both organizations have issued guidelines. The formal process, starting from the adoption of the proposal by the Commission and ending at the final adoption of the proposal as a directive by the Council, took 1¾ years for the Directive on Monitoring and Control of Large Exposures.51 The Basle Committee put forth their recommendations on the subject, Measuring and Controlling Large Credit Exposures, in October 1990. After taking account of various comments, the recommendation was reissued as a guide to best practice in January 1991.

Other Multilateral Banking Supervisory Organizations

In addition to the Basle Committee and the European Union, there are several other international banking supervisory organizations. They are the Offshore Group of Banking Supervisors; the Association of Banking Supervisory Authorities of Latin America and the Caribbean; SEANZA (Southeast Asia, New Zealand, Australia) Forum of Banking Supervisors; the GCC (Gulf Cooperation Council) Committee of Banking Supervisors; the Arab Committee on Banking Supervision; the Caribbean Group of Banking Supervisors; the Group of Banking Supervisors from Central and Eastern European Countries; the East and Southern Africa Banking Supervisors’ Group; and the West and Central African Group of Banking Supervisors. Each organization has been formed to pay special attention to the needs of its members and therefore adopts an agenda reflecting these special requirements.

Figure II.1.Decision-Making Procedures for Capital Adequacy

Multilateral Securities Firm Supervisory Organizations

There is only one primary multinational securities firm supervisory organization, the International Organization of Securities Commissions (IOSCO).52 Perhaps one reason for this striking difference is the nascent development in many countries of organized securities markets. Further, in many countries self-regulatory organizations, instead of a government-sponsored regulatory agency or commission, assume formal oversight.

IOSCO, established in 1983, has 115 regular (voting), affiliate, and associate members, who are primarily securities regulators, self-regulatory organizations, and related international organizations.53 The Technical Committee, composed of developed country members, and development committees, composed of members from countries with emerging markets, are the two principal committees through which policies or recommendations are proposed.54 Each committee has several Working Parties or Groups.

Working Parties and Groups provide their oversight committee with consensus recommendations, taking the form of a paper, guideline, or general principle. If consensus is not reached the nature of the comments on the issue are transmitted to the oversight committee. Usually approval by the oversight committee and, subsequently, the Executive Committee are pro forma when consensus is reached within the Working Party or Group. Like other organizations, IOSCO has no binding authority on its members and implementation is subject to member countries’ internal regulatory and legislative procedures.

An interesting feature of IOSCO, unlike the international banking supervisory organizations, is that it permits participation of self-regulatory organizations and other international organizations as associate and affiliate members.55 Affiliate members provide constructive input as members of the Consultative Committee, whose working groups mirror those of the Technical Committee.

Accounting Organizations

Somewhere between the public sector supervisory organizations and purely private organizations are accountancy bodies. Though technically private professional groups, many accountancy bodies can impose their standards on public companies within their jurisdiction. The two bodies most recognized as influencing banking and securities firm supervision through their accounting standards and practices are the International Accounting Standards Committee (IASC) and the Financial Accounting Standards Board (FASB) of the United States.

Established in 1973, the mission of the IASC is to establish international accounting standards. Its members consist of professional accountancy bodies, such as the American Institute of Certified Public Accountants (AICPA).56 To establish an accounting standard the IASC sets up a Task Force, which recommends a proposal to the Committee. The members vote on it (one vote per country) and it is then circulated for comment. Comments are examined by the Committee, revisions made, if necessary, and a vote is taken. If approved, the proposal becomes an accounting standard.

Not all proposals become standards, because sometimes a proposal is not acceptable to enough member countries. Thus, many IASC standards include “optional” accounting treatment of items to accommodate differences among member countries. The IASC has no binding authority, although some countries require use of IASC standards by corporations under their authority.

The United States is recognized as containing the most detailed accounting standards, and the standards of its accountancy body, the FASB, are often used as a model for IASC standards. One reason the FASB may be viewed as a leader in accounting standards is that, in the United States, the primary goal of accounting standards is to provide information to potential investors, whereas in other countries accounting standards are meant to satisfy legal or regulatory criteria, not a public disclosure requirement.

Multilateral Private Organizations

The Group of Thirty is a private, nonprofit organization aiming to “deepen understanding of international economic and financial issues, to explore the international repercussions of decisions taken in the public and private sectors, and to examine the choices available to market practitioners and to policymakers.”57 The Group sponsors symposiums and seminars and commissions monographs and reports on important topics from experts.

The Institute of International Finance (IIF) is a nonprofit, worldwide association of financial institutions.58 Members include commercial banks, investment banks, and other multinational firms and organizations. The Institute monitors global banking and financial services regulation and advances the consensus views of its members through an informal dialogue with central banks and regulatory and supervisory authorities. The Institute uses working groups and task forces to gather information and provide members with a forum for discussion of major regulatory developments.

The International Securities Market Association (ISMA) consists of member institutions from 43 countries and territories. The Association provides discussion of questions relating to international securities markets through educational seminars. It establishes uniform market practices and provides data bases and information to central monetary authorities and other institutions. ISMA uses various committees to examine issues and make recommendations on subjects of interest to its members.

Bilateral Methods of International Cooperation

Although multilateral supervisory organizations are more visible, the exchange of information among supervisors and regulators also occurs through bilateral channels. Most bilateral communication is informal, but formal bilateral agreements are well suited for some supervisory practices, potentially enhancing supervision and reducing regulatory overlap.

One common type of formal agreement is a Memorandum of Understanding (MOU), an agreement between two institutions outlining their obligations to one another. In bank and securities firm regulations, the most common MOUs provide access to official documents and information in possession of other authorities for investigatory or enforcement reasons, such as cross-border fraud. In some cases, an MOU may require affirmative action by a regulator, requiring them to report to other regulators a firm that is experiencing financial difficulties. A second type of MOU, termed a Financial Information Sharing Agreement or FISMOU, usually specifies that the parties have access to general information. FISMOUs might specify that certain information about firms operating in two jurisdictions be routinely disclosed to the regulators of both jurisdictions, such as risk assessments of related firms.

Of course, informal communication among central banks and bank supervisors continues to be the predominant form of bilateral information sharing and formal MOUs are relatively rare. For securities supervision, however, formal bilateral agreements, such as MOUs and FISMOUs, are used much more routinely. Bank secrecy laws and the relative youth of supervised securities markets, compared with banking systems, may partly account for this difference. The importance of informal bilateral communication among all bank and securities firm supervisors, however, should not be underestimated as it is the most flexible and timely, making it invaluable during crisis management.

IV Regulatory Implications of the Barings Failure

Barings plc, the oldest merchant banking group in the United Kingdom (established in 1762), was placed in “administration” by the Bank of England on February 27, 1995.59 Barings is reported to have experienced losses exceeding the entire equity capital of the firm—estimated to be $860 million at the time—from very large accumulated unhedged positions in futures contracts on the Nikkei 225 index.60 Barings’ losses apparently resulted from a $27 billion exposure that mushroomed during the three-week period ending on Friday, February 24, and especially in the last three days of this period. Barings’ position comprised relatively simple financial contracts and instruments: a $7 billion long position in exchange-traded futures contracts; a large short position in OTC Japanese government bond futures and three-month Euro-yen contracts; and significant short positions in put options and call options on the Nikkei 225 index.

While much is still not known about the Barings failure, there appears to have been a breakdown in Barings’ internal risk-management system. Barings apparently pursued a risky trading strategy that was reported, at the time, to have been undetected by Barings’ internal management control mechanisms until Friday, February 24, 1995 when it became clear that it would be unable to meet its margin requirements.61 Barings’ operations in Singapore then reported to the London operation that the bank’s derivatives contracts exposed the bank to catastrophic losses. On Friday, February 24, the management of Barings informed the Bank of England about its situation.

The collapse of Barings provides several lessons about today’s high-technology, fast-communications financial environment: (1) financial institutions can quickly become overexposed to financial risk; (2) large financial losses can result from financial positions in standard futures and options contracts, proving that it is still possible to lose money “the old-fashioned way”; (3) financial institutions need adequate management control mechanisms to assess and contain their risks; (4) financial institutions can be allowed to fail without bailouts using public funds—as an essential part of the mechanism of market discipline—at least for moderate-sized institutions when the condition of other institutions is not suspect; and (5) financial markets are truly global—actions by a single Singapore-based trader of a British merchant bank can have pronounced effects in markets around the globe.

The collapse of Barings has come at a time when regulators are considering a “new approach to regulatory capital requirements,” which permits money center banks to rely on their internal risk-management models to determine how much capital to hold against OTC transactions. In addition, the systemic implications of Barings’ failure were more easily contained partly because their positions on the exchanges had been marked-to-market daily. This experience demonstrates that risk management in futures exchanges is generally more reliable than risk management in the OTC derivative markets operated by international banks.

The Barings failure raises questions about how managers of financial institutions and regulators together can prevent similar events in the future. The crisis points to three important issues of concern: (1) the effectiveness of internal management-control systems and risk management; (2) the quality of risk management, surveillance, and coordination by financial exchanges; and (3) the important roles, and the coordination, of home and host country supervisors in detecting risks for individual institutions and in preventing systemic risks.

Internal Risk Management

Current information suggests that the Barings failure was brought on by the absence of adequate internal audit and management-control systems. Because the losses occurred in exchange-traded markets, where positions are marked-to-market, variation margin payments to the various exchanges occurred daily. At least one person at Barings was responsible for making these daily payments, and it has been alleged that the same person who was placing the trades was also responsible, at least in part, for settling the daily margin payments. If a trader at Barings could make margin payments and enter trades into the computer systems without independent verification then there were opportunities to falsify or conceal positions. If this is what occurred, then Barings was lacking appropriate checks and balances that are normally secured by dividing responsibilities among different levels of management. The separation of the responsibility for payments and settlements from trading functions and independent trade verification are important elements of an effective risk-management system.

A broader implication is that operational risk (the risk due to mismanagement, trading errors, imprudent trading strategies, and so on) should receive greater attention in risk-management systems by all parties involved, including financial institutions, exchanges, and supervisors. Formal risk-management models constitute one part of a comprehensive risk-management process that includes independent risk control and management, extensive auditing and verification procedures, and formal position limits. Effective management and control requires that senior management understand the derivatives positions taken by the firm.

Assuming Barings utilized the latest risk-management technologies, what caused the Barings model, or its other internal controls, to fail to detect that the value-at-risk from positions taken potentially exceeded the capital of the firm? Why were these risky positions not detected by host or home country regulators? If a “rogue” trader can put at risk the entire capital of a merchant bank like Barings, then to what extent should banks be permitted to determine regulatory capital requirements on market risk based on their own internal model? In light of the Barings failure, should the Basle Committee on Banking Supervision rethink its current proposals to allow banks to use their internal risk-management models to calculate regulatory capital requirements?

Risk Management on Organized Exchanges

Because Barings’ large unhedged position resulted from the accumulation of exchange-traded contracts, the quality of risk management at the exchanges comes into question. The exchanges involved are the Singapore International Monetary Exchange (SIMEX), the Osaka Securities Exchange (OSE), and the Tokyo International Financial Futures Exchange (TIFFE). The SIMEX continued to collect margin deposits from Barings until Friday, February 24. On Monday, SIMEX doubled the requirements of both initial and variation margins on the Nikkei 225 futures contract. Some clearing members and customers expressed concern about the ability of SIMEX to continue to guarantee settlement of payments, and the Singapore Monetary Authority immediately issued a press release agreeing to stand behind the SIMEX clearinghouse. While the Monetary Authority’s guarantee is important to quell concerns of market participants, the exchange’s clearinghouse has procedures to ensure payment should a clearing member default. SIMEX, like most exchanges, has a “guarantee fund” or “compensation fund” providing a method for collecting monies from exchange members in case a defaulting clearing member’s losses exceeds clearinghouse funds.

SIMEX has a large trader-reporting system whereby accounts with more than 100 Nikkei 225 futures contracts are reported to their surveillance authorities. House (the firm’s proprietary account) and customer accounts are separated, so in principle, the exchange and other authorities can view the trades of the proprietary trading arm of the firm. It is possible that viewed from SIMEX, the trader was not taking abnormal positions. Apparently, the Barings’ operation in Singapore frequently carried large positions, which were offset by the opposite positions in similar contracts on the Osaka and Tokyo exchanges, in an attempt to profit from the small differences between the prices of like contracts. This may have entitled Barings to a “hedge exemption” from speculative position limits, which typically is granted to nonspeculators holding “hedged” positions. If so, then SIMEX may not have noticed an abnormal pattern associated with the size of the position on the exchange.62 The implication is that SIMEX may have observed Barings’ large proprietary position in advance of the defaulted margin call, but it may not have viewed the position as a problem. Alternatively, there is some speculation that Barings “hid” large trades by creating fictitious customer accounts, none of which appeared to have excessively large positions. Assuming SIMEX was able to observe Barings’ large positions and viewed them as hedged positions, better international coordination among exchange surveillance authorities or more frequent verification of the associated hedged positions might have detected a problem.

Another important issue with potential systemic implications is the efficiency and speed with which positions of an insolvent entity are unwound, and the distribution of losses, if any, among counterparties. If a customer of a firm in default cannot gain access to its own funds, then that customer may not be able to make payments to its counterparties in other unrelated transactions. In principle, this could lead to the failure of otherwise solvent entities.

The contracts of an insolvent financial institution are more easily resolved, and expose the financial system to fewer systemic problems, if they are exchange traded. On Monday, February 27, all of Barings’ accounts on the three exchanges were suspended; the accounts were quickly freed on Tuesday, February 28. Customer accounts with Barings were then transferred to another clearing firm, while Barings’ proprietary accounts are being handled by the “administration” process. SIMEX separates clearing members’ customer and proprietary accounts, making it relatively easy to transfer customer positions without customers incurring losses due to Barings’ proprietary trading. There was, however, some temporary uncertainty over customers’ access to their accounts due to reports that Barings had not maintained this strict separation of proprietary and customer accounting. In contrast, neither the OSE or TIFFE separated customer and proprietary trading accounts. As a result, the separation of Barings’ proprietary trading from that of its own customers is taking some time to sort out. This could potentially pose difficulties for customers and their counterparties. On the whole, however, the relative ease with which Barings’ default is being managed might be taken as evidence that the clearinghouse mechanism is useful for reducing systemic risk. In contrast, the resolution of positions taken by an insolvent entity, such as Barings, is less straightforward on over-the-counter contracts, and will take considerably more time to unwind. The identity of the over-the-counter counterparties, the size of their exposures, and how these contracts will be resolved under the process of U.K. “administration,” is currently unknown.

Overall, the futures exchanges in Osaka and Singapore operated effectively and efficiently and have retained their capital and reputations. Daily marking-to-market of positions, combined with initial margins—keyed to statistical estimates of price volatility—provided a sufficient cushion to deal even with this extraordinary situation. Nevertheless, several improvements are possible. First, Barings had been able to borrow securities—unsecured—from Japanese financial institutions, in the informal Tokyo securities repurchase market, to meet margin calls at Osaka. These securities were liquidated by the Osaka exchange and lost to the original owners. One way to reduce risks in the future would be to formalize the securities repo market into a collateralized lending facility. Second, the legal claim to margin money under national bankruptcy laws will need to be clarified. Third, the loss-sharing rules of the futures exchanges, particularly SIMEX, will also need to be strengthened and made transparent.

V Increasing Importance of Institutional Investors

One of the most significant recent developments in international financial markets is that individual investors have increasingly delegated the management of their portfolios to professional fund managers. The consequence of this is that the investor base in securities markets in industrial countries, and increasingly in developing countries, is dominated by a relatively small number of large institutional investors. In addition, the international diversification of institutional portfolios has developed in tandem with the institutionalization of savings and portfolio management. This trend toward the international diversification of institutional portfolios has increased the sensitivity of securities markets—especially the smaller, less liquid markets, notably in developing countries—to the behavior of a relatively small number of investors.

Size of Institutional Investors’ Portfolios

The institutional investor community can be defined broadly to include public and private pension funds, life and other insurance companies, mutual (open-end) funds, closed end funds, hedge funds, trusts, foundations, endowments, and proprietary trading by investment banks, commercial banks, and securities companies. To illustrate the growing importance of these investors, total assets of the 300 largest U.S. institutional investors rose from 30 percent of GDP in 1975 ($535 billion) to more than 110 percent of GDP in 1993 ($7.2 trillion).63 Similar changes in total assets under management are recorded for other industrial countries.

Pension funds, insurance companies, and mutual funds in five major industrial countries had close to $13 trillion in assets under management in 1993.64 To put this number in perspective, the global equity market capitalization in the same year was $14.1 trillion, and the outstanding stock of government debt for the seven largest industrial countries was $9 trillion.65 The importance of institutional investors has increased markedly: since 1980, the assets managed by the institutional investors in five of the major industrial countries have increased by more than 400 percent and have more than doubled as a percent of GDP.

U.S. institutional investors control the largest pool of assets by a significant margin—U.S. pension funds, insurance companies, and mutual funds alone managed more than $8 trillion of assets in 1993 (Table II.6). U.S. pension funds have maintained their status as the largest institutional investor in the world. Insurance companies in all industrial countries control sizable asset pools, with Japanese life insurers being especially notable with $1.48 trillion of assets in 1993. Expressing managed assets as a percent of GDP, the United Kingdom takes the lead with professionally managed assets representing 165 percent of GDP in 1993. Institutional investors in the United States also control assets in excess of annual GDP.

Table II.6.Assets of Institutional Investors(In billions of U.S. dollars)
198019881990199119921993
Pension funds
Canada43.3131.3171.8188.4191.7
Germany17.241.655.258.662.653.5
Japan24.3134.1158.8182.3191.9
United Kingdom151.3483.9583.6642.9670.5695.7
United States667.71,919.22,257.33,070.93,334.33,571.4
Life insurance companies
Canada36.885.5106.1118.1131.8132.7
Germany88.4213.5299.5325.7341.4354.3
Japan124.6734.6946.91,113.71,214.81,476.5
United Kingdom145.7358.9447.9516.7574.7619.3
United States464.21,132.71,367.41,505.31,624.51,784.9
Non-life insurance companies
Canada9.222.726.823.3
Germany36.787.6126.3127.4187.9198.8
Japan34.6156.1190.3215.5218.5167.9
United Kingdom31.372.285.289.695.697.1
United States182.1453.9529.2591.6628.7640.8
Mutual funds1
Canada3.917.521.543.252.986.7
Germany222.499.9145.5166.2171.6205.2
Japan360.8433.9353.5323.9346.9448.7
United Kingdom16.876.791.5104.491.2141.3
United States292.9810.31,066.91,348.21,595.42,011.3
Total
Canada93.2257.0326.2373.0376.4
Germany164.7442.6626.5677.9763.5811.8
Japan244.31,458.71,649.51,835.41,972.1
United Kingdom345.1991.71,208.21,353.61,432.01,553.4
United States1,606.94,316.15,220.86,516.07,182.98,008.4
Total (in percent of GDP)
Canada35.252.256.863.366.1
Germany20.337.141.742.742.747.4
Japan23.150.356.354.853.8
United Kingdom64.1118.3123.5133.8137.1165.3
United States59.388.194.5113.9119.0125.6
Sources: Bank of Canada, Bank of Canada Review, various issues; Bank of Japan, Economic Statistics Monthly, various issues; Chuhan (1994); International Monetary Fund, International Financial Statistics; Investment Company Institute; United Kingdom, Central Statistical Office, Financial Statistics, various issues; United States, Board of Governors of the Federal Reserve System, Flow of Funds Accounts, various issues; and IMF staff estimates.

The numbers in the first column are for 1983, except for Canada.

Public and special funds.

Investment trusts.

Sources: Bank of Canada, Bank of Canada Review, various issues; Bank of Japan, Economic Statistics Monthly, various issues; Chuhan (1994); International Monetary Fund, International Financial Statistics; Investment Company Institute; United Kingdom, Central Statistical Office, Financial Statistics, various issues; United States, Board of Governors of the Federal Reserve System, Flow of Funds Accounts, various issues; and IMF staff estimates.

The numbers in the first column are for 1983, except for Canada.

Public and special funds.

Investment trusts.

Pension funds and insurance companies have traditionally been the most important institutional players in industrial country financial markets. Although pension funds and insurance companies still control sizable (and growing) portfolios of securities, an important factor in the institutionalization of savings has been the rapid growth of mutual funds and closed end investment companies. Mutual funds alone accounted for about $3 trillion in 1993 of private wealth. Although this sum is not as large (yet) as managed assets of insurance companies and pension funds, mutual fund assets have increased at a much faster pace than have the assets of other institutional investors. Some of the important factors underpinning this growth are the increased sophistication of individual investors, technological improvements in information transmission and clearance and settlement of securities, and increased emphasis by industrial countries on capital account convertibility (and, consequently, the integration of capital markets).

Hedge funds are important institutional investors, but it is difficult to obtain comprehensive data on the majority of these private funds; their activities are therefore not included in the figures reported in Table II.6. There is no universally accepted definition of a “hedge fund,” but two important features are that they are unregulated and are often highly leveraged. Most hedge funds appear to be located in the United States.66 The reason that they are unregulated is because either they have fewer than 100 investors, and thus do not have to comply with SEC disclosure and registration requirements (as stipulated by the Investment Company Act of 1940), or they are domiciled offshore. Onshore funds are structured as investment partnerships and the minimum investment is typically in the range of $350,000–10,000,000. The number of investors is generally much greater and the minimum investment much smaller for offshore funds, which are often structured simply as open-end mutual funds, albeit with higher minimum investment than most retail mutual funds.

Hedge funds have been in existence since the 1940s, but first became prominent in the 1960s. The number of onshore hedge funds is not known, but most estimates are in the neighborhood of one thousand, up from around one hundred in 1987; assets under management have doubled since 1991.67 The vast majority of hedge funds have capital below $100 million, and only about a dozen have assets currently exceeding $1 billion. The largest hedge funds had assets between $6 billion and $10 billion in 1994. Although total assets of all hedge funds are estimated to be around $75–100 billion, the possible positions taken by hedge funds can be much larger. This is because these investors are not constrained by leverage restrictions on investment enterprises that fall under the purview of the Investment Company Act of 1940. In fact, some hedge funds committed to a particular investment opportunity may be leveraged between 5 and 20 times their capital. Factoring leverage into the net capital of hedge funds leads to the conclusion that they are a potentially important player in global capital markets.

Although there are hundreds of hedge funds, they differ markedly by investment objective. The lesser known “traditional” hedge funds are often interested in cross-security arbitrage within an equity market. The recently more visible hedge funds—commonly referred to as “macro,” “opportunistic,” or “directional” hedge funds—are chiefly interested in the currency and bond markets, typically taking highly speculative, and highly leveraged positions through liberal use of bank loans, options, futures, and other derivatives. Although it is estimated that there are only 15 or so of these macro hedge funds, they appear to control a very significant portion of the industry’s assets.68

There are several regulatory considerations associated with the rise in the importance of institutional investors. As households increasingly delegate their investment management decisions to professional fund managers, the effective investor base is changing from a very large number of small investors toward a relatively small number of large investors. With greater concentration of wealth in the hands of professional fund managers, financial markets must cope with the effects of the attendant increase in the market power of market participants. Chief among these effects is the increased likelihood of market manipulation, and even less efficient markets.69 Furthermore, as the investor base becomes more highly concentrated, the likelihood of coordination failures may increase and this can produce abrupt changes in market liquidity.70 Market manipulation can be especially important in smaller markets, notably the emerging markets.

Hedge funds raise some unique regulatory considerations because of the unusually high degree of leverage. Hedge funds use banks for a variety of services, but the services that raise regulatory considerations are those that create credit risk to the banking system—foreign exchange trading lines, repo lines, and loans. Foreign exchange trading lines and repurchase lines seem to be key services provided to hedge funds by banks in the United States.71 But this is not surprising as banks are the principal providers of these services to all types of institutional investors. In fact, hedge funds appear to account for a relatively small portion of the total services provided by U.S. banks to institutional investors; however, care should be taken in drawing inferences about the exposure of any one bank when studying average exposures of all banks. An important consideration, therefore, is that individual banks have effective risk-management systems for all customers.

Are Institutional Investors Internationally Diversified?

In light of the substantial growth of assets under institutional management, an important question for international finance is the degree to which institutional investors have diversified their portfolios internationally. The evidence points to the accumulation of a substantial amount of foreign assets by institutional investors. The share of portfolio assets that are foreign securities also has increased for some institutional investors (Table II.7). Specifically, the foreign asset shares of pension funds have been on a gradual upward path at least since 1980, and in 1993 ranged from a low of 4.5 percent for German pension funds to a high of almost 20 percent for U.K. pension funds.

Table II.7.Institutional Investors’ Holdings of Foreign Securities(In percent of total assets)
198019881990199119921993
Pension funds
Canada4.15.35.88.510.210.3
Germany3.84.54.54.34.5
Japan0.56.37.28.48.49.0
United Kingdom10.116.518.020.822.019.7
United States0.72.74.24.14.65.7
Life insurance companies
Canada3.31.91.61.92.31.8
Germany0.60.61.01.0
Japan2.714.213.512.511.49.0
United Kingdom5.59.510.812.412.711.6
United States4.13.63.63.63.7
Mutual funds
Canada19.919.517.516.216.717.1
Germany56.353.547.645.2
Japan19.17.913.09.9
United Kingdom37.139.237.936.0
United States6.610.1
Sources: Bank of Canada, Bank of Canada Review, various issues; Bank of Japan, Economic Statistics Monthly, various issues; Bisignano (1994); Chuhan (1994); European Federation of Investment Funds and Companies; International Monetary Fund, International Financial Statistics; InterSec Research Corporation; United Kingdom, Central Statistical Office, Financial Statistics, various issues; United States, Board of Governors of the Federal Reserve System, Flow of Funds Accounts, various issues; and IMF staff estimates.

Investment trusts.

Sources: Bank of Canada, Bank of Canada Review, various issues; Bank of Japan, Economic Statistics Monthly, various issues; Bisignano (1994); Chuhan (1994); European Federation of Investment Funds and Companies; International Monetary Fund, International Financial Statistics; InterSec Research Corporation; United Kingdom, Central Statistical Office, Financial Statistics, various issues; United States, Board of Governors of the Federal Reserve System, Flow of Funds Accounts, various issues; and IMF staff estimates.

Investment trusts.

While there seems to be a clear but gradual trend toward internationally diversified portfolios of pension funds, the behavior of insurance companies is less clear. Insurance companies have in general not increased their international diversification over the same period, and in several countries even show some decrease in the 1990s. Insurance companies also are not as internationally diversified as pension funds, with the possible exception of Japan. Developments in Canada illustrate the more general trends in the period 1980–93: foreign asset shares of insurance companies are only about one sixth the size of pension funds, and, while pension funds have more than doubled their foreign asset share, insurance companies have cut their share by almost half.

The aggregate portfolio of mutual funds has several interesting features. First, mutual funds in Canada, Germany, and the United Kingdom are far more internationally diversified than U.S. or Japanese mutual funds. Second, U.S. mutual funds stand alone as displaying a clear trend toward increased international diversification. In fact, mutual funds in the other countries appear to have either leveled off in terms of foreign portfolio shares or to have decreased their share of foreign assets.72 Third, mutual funds in most countries (with the exception of Japan) are significantly more internationally diversified than other institutional investors. Even in the United States, where mutual funds are among the least internationally diversified of all countries, their foreign asset holdings are roughly double those of insurance companies and pension funds, as a share of total assets.

Despite the general trend toward international diversification, especially by mutual funds and pension funds, it is also clear that this trend in international diversification of portfolios is overshadowed by the small share of foreign securities in institutional portfolios, especially for insurance companies and pension funds. A well-known rule of thumb from modern portfolio theory is that an optimally diversified portfolio for an individual investor should have country weights corresponding to the ratio of a country’s market capitalization to the world market capitalization.73 Using the market capitalization measures of the International Finance Corporation, this “world market portfolio” in 1993 would have had 37 percent of its investments in the United States, 21 percent in Japan, 8 percent in the United Kingdom, 22 percent in other industrial countries, and just under 12 percent in the emerging markets. Although, the degree of international diversification by aggregated mutual funds is greater than some other institutional investors, all types of institutional investors are much less internationally diversified than this world market portfolio.74

Although the share of foreign assets in the portfolios of some institutional investors is quite low relative to standard benchmarks, the level of cross-border securities holdings is substantial. For example, in 1993, a 5.7 percent foreign asset share in U.S. pension funds translated into foreign security holdings by U.S. pension funds of $203.6 billion. With U.S. mutual funds and insurance companies together holding roughly equivalent amounts, these three types of institutional investors in the United States alone control about $400 billion in foreign securities. Even though cross-border holdings seem to be far from an optimally diversified portfolio, these holdings could be an important source of funds.

One explanation for the low share of foreign securities in institutional portfolios is that the seeds of international portfolio diversification have only recently been planted and institutional investors are poised to make important advances in international diversification by the end of the century. For instance, estimates put U.S. private pension funds’ foreign holdings at 9.5 percent of assets by year-end 1995, representing a more than 100 percent increase in three years.75 These estimates are underscored by a recent poll of U.S. pension funds, which found that during 1995, 51.3 percent of them plan to increase their foreign equity holdings, 21.3 percent plan to increase their foreign fixed income holdings, and only about 2.0 percent plan to reduce either one.76

A second reason for the heavy bias of institutional investors’ portfolios toward domestic assets is that it reflects the well-known tendency of individual investors in all industrial countries to display a very marked “home-asset preference.”77 Many reasons have been put forward to explain this preference, including transactions costs, foreign exchange risk, uncertainties about expected returns, and unfamiliarity with foreign markets and tax laws. Institutional investors report that this aversion is an important consideration. Specifically, when asked why foreign asset holdings are such a small fraction of total assets, some fund managers report that the underlying investors or trustees of the funds they manage are “very conservative.” This aversion to foreign markets seems to be largely a reflection of a lack of familiarity with foreign economies and financial markets.

A third possible reason for the limited degree of international diversification by institutional investors in industrial countries is that fund managers face binding, externally imposed, explicit quantitative constraints on foreign asset holdings. These types of constraints can arise from two sources. First, internal committees that set broad investment criteria for fund managers may restrict or even prohibit the fund from investing in foreign assets. In this case, the explanation for the observed low share of foreign asset holdings requires looking beyond the preferences of a fund’s manager. The second reason is that government regulations of foreign asset holdings of institutional investors appear to be restrictive for at least some institutions in some countries (Table II.8). For example, a recent poll of Canadian pension fund managers found that 88 percent of them would increase foreign asset holdings in their portfolios above the 20 percent government-imposed ceiling if they could, with most favoring a 25–35 percent foreign asset allocation; only 18 percent would raise foreign holdings above 35 percent.78 However, it is also reported that government-imposed restrictions on foreign asset holdings are not binding for many institutional investors in industrial countries.79 This suggests that regulatory constraints on foreign asset holdings may explain some of the difference between institutional portfolios and the predictions of standard portfolio theory, but they do not fully account for investors’ preference for home assets.

Table II.8.Regulatory Constraints on Outward Portfolio Investment of Institutional Investors in Selected Industrial Countries1
Country/RegionPension FundsInsurance CompaniesMutual Funds
CanadaA December 1991 law progressively raised the ceiling on foreign investment from 10 percent to 20 percent in 1994.A June 1992 regulation removed ceilings on foreign investments, but limits may be imposed based on prudential considerations.Limit of 20 percent on foreign assets in the Registered Retirement Savings Plans (RRSP)-eligible funds.
FranceAt least 50 percent of assets must be invested in securities guaranteed by the state.Investments are subject to the matching assets rule; the location rule; and the allocation of assets rule.Subject to disclosure and asset diversification rules. A fund may not hold more than 10 percent of any one category of securities of one issuer.
GermanyFive percent of the assets of the technical provision fund and 20 percent of the other restricted assets in respect of business written in European Economic Area (EEA) States may be localized outside the EEA States. No restriction for free assets. Matching rules apply.Five percent of the assets of the technical provision fund and 20 percent of the other restricted assets in respect of business written in European Economic Area (EEA) States may be localized outside the EEA States. No restriction for free assets. Matching rules apply.None.
JapanPrivate funds are subject to 30 percent foreign asset limit. Fifty percent of assets must be in guaranteed fixed return, domestic yen vehicles.Holding of securities issued by nonresidents is limited to 30 percent of total assets; the same ratio applies to purchases of foreign currency denominated assets.
United KingdomNone.Subject to matching and localization rules, which require them roughly to balance liabilities expressed in a particular currency with assets in the currency. A company must ensure that its liabilities are covered by assets of appropriate safety, yield, and marketability, having regard to the classes of business carried on, and that its investments are appropriately diversified and adequately spread and that excessive reliance is not placed on investments of any particular category or description.Collective investment schemes (unit trusts) are required to invest at least 90 percent of their assets in transferable securities in markets, selected by the fund manager in consultation with the trustees, which are regulated, recognized, operate regularly, and are open to the public.
United StatesRegulated by a special federal law—Employee Retirement Income Security Act (ERISA). Permissible investments subject to the “prudent expert” rule, which includes a requirement to give consideration to diversification and liquidity factors. Otherwise no explicit restrictions on holding foreign securities, including foreign equities and foreign currency denominated bonds.U.S. state insurance regulations attempt “to prevent or correct undue concentration of investment by type and issue and unreasonable mismatching of maturities of assets and liabilities.” These laws usually allow an unrestricted “basket” of investments for certain amount of assets, which can be allocated to foreign securities in the range 0–10 percent of total assets.Primarily regulated by the U.S. Securities and Exchange Commission (SEC) under federal laws. An open-ended fund may not hold more than 15 percent of its net assets in illiquid assets. Otherwise no explicit restrictions are imposed on investment in foreign securities.
European UnionThe EC life and non-life insurance directives intend to remove all legal barriers for the creation of a common market in insurance. They also set out provisions to harmonize rules on admissible investment.None.
Sources: Chuhan (1994); International Monetary Fund (1993); Organization for Economic Cooperation and Development (OECD); and national authorities.

For the securities houses of these countries, there are no explicit regulatory restrictions on foreign exchange positions and other cross-border investments.

Sources: Chuhan (1994); International Monetary Fund (1993); Organization for Economic Cooperation and Development (OECD); and national authorities.

For the securities houses of these countries, there are no explicit regulatory restrictions on foreign exchange positions and other cross-border investments.

If government-imposed constraints on many institutional investors’ foreign asset holdings are not important or binding, then there would appear to be a preference for home assets by all institutional investors, and a much more marked preference for home assets by pension funds and insurance companies than by mutual funds. The fact that these differences exist may point to the influence of more considerations in portfolio management decisions than elementary portfolio theory suggests. Whereas the share of foreign assets in mutual funds probably reflects fairly accurately the preferences of the underlying investors in the long run, this may not be a valid conclusion about pension funds and insurance companies.

One reason that pension funds and insurance companies may display a preference for domestic assets is that pension fund trustees or investment management companies may display low-risk tolerance. Also important is the fact that most of the underlying investors are typically quite far removed from portfolio management decisions. A key reason why trustees or investment management companies might display lower risk tolerance than underlying claimants on the pension fund or insurance premium fund is shortfall risk. If the investment manager, or company, bears more downside risk than the underlying investors and, in addition, does not capture fully the upside potential, then the manager will optimally specialize the portfolio in safe, domestic assets.80 In effect, contracts to manage asset pools for pensions or insurance companies typically generate a high degree of risk aversion on the part of fund managers because they do not share risk optimally.81 These same contract-induced biases toward home assets may not be present in mutual funds because, given the mandate of a particular mutual fund (e.g., emerging markets), managers are typically compensated as a proportion of net asset value; this type of contract tends to share risk more efficiently between the fund manager or investment company, or both, and the underlying investors. Another key difference between mutual and defined benefit pension funds is that mutual funds are by definition fully funded: unlike a pension fund, there is no risk of a mismatch in its assets and liabilities. Further, in the case of mutual funds, it is easier for underlying investors to signal their investment preferences simply through fund selection.82

Institutional Investment in Emerging Markets

In 1993, the capitalized value of equity securities in emerging markets represented slightly less than 12 percent of the capitalized value of all equity markets.83 Even though it is not possible to provide a country breakdown of foreign asset holdings of institutional investors, on average they hold substantially less than 12 percent of their total assets in emerging markets securities.

An indirect measure of the degree to which institutional investors have increased their holdings of emerging markets securities is provided by data on portfolio flows from industrial countries to emerging markets. These data represent a reasonably good measure because, although gross portfolio outflows from industrial countries include some retail cross-border transactions (i.e., by individuals), most of the portfolio inflows to emerging markets are flows from institutional investors.84 This suggests that the ratio of total portfolio inflows to emerging markets to outflows from all industrial countries provides a measure of the fraction of institutional flows to foreign markets that are targeted to emerging markets.

According to these ratios, the share of institutional investment in emerging markets appears to have increased at a rapid pace (Table II.9). Interpreted at face value, these figures suggest that, in 1993, more than $16 out of each new $100 foreign investment went to emerging markets in 1993, up from about $0.50 in 1987. In addition, securities originating in the Western Hemisphere accounted for a large share of total inflows; Mexico alone accounted for half of these flows.85

Table II.9.Industrialized Country Securities Investment Flows in Emerging Markets(In percent of foreign securities investment flows)
1987198819891990199119921993
Africa−0.6−0.1−0.1−0.21.1−0.1
Asia1.60.20.80.61.52.32.7
Europe0.20.60.60.40.31.01.7
Middle East0.32.20.40.40.30.20.5
Western Hemisphere−1.00.2−0.410.78.110.211.4
Mexico−0.80.50.12.04.25.65.7
All emerging markets0.43.01.211.910.214.816.3
Memorandum items
Outflows from industrialized countries (in billions of U.S. dollars)123.4207.5276.4170.1306.8320.3495.3
Emerging markets capitalization as a share of world capitalization (in percent)4.15.06.36.57.58.811.6
Mexican markets capitalization as a share of world capitalization (in percent)0.10.10.20.30.91.31.4
Sources: International Finance Corporation, Emerging Stock Markets Factbook, various issues; and International Monetary Fund, Balance of Payments Statistics Yearbook 1994, Part 2; and IMF staff estimates.
Sources: International Finance Corporation, Emerging Stock Markets Factbook, various issues; and International Monetary Fund, Balance of Payments Statistics Yearbook 1994, Part 2; and IMF staff estimates.

It is conventional wisdom that the source of portfolio flows to emerging markets in the Western Hemisphere is primarily U.S. institutional investors, especially mutual funds. In contrast, the investor base for Asian emerging markets is more diverse, including institutional investors from Europe, the United States, and Japan. As a share of total funds invested, the amount of foreign securities held by institutional investors in the United States is smaller than the amount held by institutional investors in other countries. However, the sheer size of U.S. institutions means that the dollar value of U.S. institutional investments in emerging markets, and especially to countries in the Western Hemisphere, is quite substantial.

It is estimated that in 1993, U.S. institutional investors owned about 30 percent of the $425 billion in debt outstanding issued in emerging markets, including $80 billion in Brady bonds, $254 billion in bank debt, and $91 billion in Eurobonds, global bonds, and yankee bonds.86 This proportion is in line with an IFC estimate that between 20 and 50 percent of all net inflows to emerging markets originated in the United States. This interest in emerging markets by U.S. institutional investors has been spearheaded by mutual funds. It is estimated that, in 1993, U.S. mutual funds invested the net amount of $20 billion in emerging markets, bringing their emerging market holdings to $100 billion at the end of 1993.87 This translates into an emerging markets position in the neighborhood of 5 percent of total assets, and almost half of total foreign assets. Thus, U.S. mutual funds have significantly diversified their portfolios into emerging markets.

Closed end investment companies have also made significant investments in the emerging markets. The structure of closed end investment companies makes them an especially good vehicle for investing in less liquid markets. As a result, closed end funds are a relatively more important source of investment in emerging markets than in industrial countries.88 It is estimated that, in September 1994, emerging markets assets held by U.S. and overseas closed end funds totaled $45 billion.89

The structure of closed end funds facilitates longer-term investments and therefore may be relatively more attractive to developing countries than some other portfolio flows. Specifically, portfolio weights in closed end funds are, on average, not nearly as sensitive to market volatility changes as open-end funds. One reason is that closed end fund managers are shielded from actual and expected redemptions because the claims on the pool of assets are traded on a stock exchange. In contrast, open-end funds are required to redeem claims on demand, which can have an important effect on portfolio decisions.90 “Redemption risk” is probably one of the most important reasons for the much higher turnover ratios of open-end funds (on average) than closed end funds.91 The consequence is that open-end funds tend to concentrate on an emerging market’s larger firms because they often have the most liquid securities. Closed end funds, on the other hand, not only do not face redemption risk, but they are able to select lesser known, less liquid, securities. Because the securities of the largest firms in emerging markets are often directly available to individual investors through ADRs and GDRs or direct transactions in that country, this puts open-end funds at a disadvantage.92

In 1994 and early 1995, U.S. open-end and closed end funds gave rise to substantial net inflows to emerging markets. Although it is well known that net contributions to U.S. mutual funds have slowed since 1993, contributions to international mutual funds that have investment mandates for emerging markets were strong for most of 1994 (Table II.10). Moreover, net contributions to emerging markets in the Western Hemisphere equity funds actually rebounded in early 1995 from the last few months of 1994, despite the Mexican crisis. Contributions in January 1995, however, were only a small fraction of those one year earlier. In addition, redemptions from bond funds continued to advance rapidly in early 1995.

Table II.10.International Mutual Funds Based in the United States1(Net fund inflows monthly in millions of U.S. doll
Jan.

1994
Feb.

1994
Mar.

1994
Apr.

1994
May

1994
June

1994
July

1994
Aug.

1994
Sept.

1994
Oct.

1994
Nov.

1994
Dec.

1994
Jan.

1995
(Equity)
International funds2
Growth3,6262,4522,3531,6101,5069271,7062,2511,1261,362299359539
Emerging market equity1,398830−150455346157224688411341−51−5889
Total return495352208125173−22891641768898−9424
Latin American equity858649−256−61355465305111110−24−8021
Single country equity6490803−230−220913467−6−2−14
Small cap299397862762210566654117−23−20
Chinese equity75855715231210−413−20−26
Canadian equity471819−7−188672019−2−13−2−35
Japanese equity182521639716329559−127−1071−47−32−55
European equity43537225119257−123103152−182−44−128−108−109
Pacific equity with Japan4022842417021218717127726160−39−27−74
Pacific equity without Japan1241379522112932117227352156−116−49−83
Global funds
Growth1,4681,338704789771459702987679703252373259
Small company28519312810410580932108776602878
Total return32026319812110276781227862532627
Equity sector3362411024949249271704512639−27−9
Asset allocation8595744792084313052133231564056−114−30
Total11,2148,7274,9594,1144,4832,7983,8645,8573,1853,282664151583
(Taxable Bond)
International funds
Single country bond−1−1−1−1−1−1−1−1
Emerging market bond14274−21107270−20115714−33−146−47−40
Global funds
Government bond79367−12−43−30−69−32−81−103−9232−88−101
North American bond554313−286−335−117−77−130−87−180−156−77−442−237
Short bonds−218−388−509−217−340−448−305−381−370−295−314−311−329
Bond (general)420136−39−18123−119−213−179−289−196−325−233−374
Total976502−866−507−94−733−668−671−928−773−830−1,123−1,081
Grand total (equity plus taxable bond)12,1919,2294,0933,6074,3892,0663,1955,1862,2572,509−166−971−498
Source: Strategic Insight Simfund.

Open- and closed end funds.

International funds invest in non-U.S. securities only; global funds invest in foreign and U.S. securities.

Source: Strategic Insight Simfund.

Open- and closed end funds.

International funds invest in non-U.S. securities only; global funds invest in foreign and U.S. securities.

In contrast to the appetite of U.S. mutual funds for emerging markets securities, U.S. pension funds appear to show limited interest in emerging markets. Emerging markets mandates for many pension funds in the past couple of years have been well below 1 percent of total assets, and average emerging markets holdings in 1993 are estimated at about 0.5 percent of total assets. This estimate places pension fund holdings of emerging markets securities at about $18 billion in 1993, less than one fifth of the holdings of U.S. mutual funds with emerging markets mandates in 1993 (Table II.6).93

There has been an important shift in recent years toward diversification into emerging markets by some institutional investors. However, as a percent of total assets, institutional investors are a long way from an emerging markets portfolio share representative of emerging markets capitalization relative to world market capitalization. Most estimates place the average share of emerging markets securities in institutional investors’ portfolios around 1 percent. This low weight of emerging markets is consistent with the strong bias of pension funds and insurance companies away from foreign markets in general, and emerging markets specifically. There seems to be an even stronger aversion of the more “conservative” types of institutional investors to emerging markets than to foreign markets. For example, a recent poll of U.S. pension funds indicated that 7 percent of them would allocate no additional contributions to emerging markets in 1995, 27 percent would allocate between 1 and 5 percent of new contributions, and only 2 percent would allocate more than 4 percent of new contributions to emerging markets.94

Despite these relatively small emerging markets portfolio weights of institutional investors, two facts are noteworthy. First, with an estimated 1 percent of assets invested in emerging markets, the share of emerging markets holdings in total foreign holdings is therefore above 10 percent (from Table II.7) for many institutional investors. This might indicate that investors’ aversion is focused on foreign investments in general, not necessarily emerging market investments. Second, since 1990, the share of new foreign portfolio investment that has been targeted to emerging markets appears to have exceeded the market capitalization share. It is noteworthy that these general trends seem to be much more marked for Mexico, suggesting possibly an overweighting of Mexican securities.

Instead of interpreting the emerging markets portfolio weights of institutional investors in light of modern portfolio theory, one might instead question the benchmark that is used to gauge how far institutional investors have moved toward “optimal” international diversification. Portfolio managers from a variety of institutions report that the relevant benchmark for them is not a market-capitalization weighted portfolio, but instead a much narrower benchmark, such as a domestic stock index for mutual funds or an average of other institutional investors’ portfolios for pensions, endowments, and trusts. Furthermore, the market-capitalization benchmark can be criticized for ignoring cross-country differences in liquidity, information, accounting practices, the market capitalization of traded firms as a percentage of market capitalization of all domestic firms, transactions costs, and custody and settlement systems. This seems especially relevant in light of the fact that institutional investors report that these factors—especially liquidity—are a fundamental reason why they hold such a small fraction of emerging markets securities. Even pension-fund and trust-portfolio managers, which, on the surface, might not appear to have the same concerns about liquidity as mutual funds (e.g., redemptions), state that annual turnover of their portfolios is often quite high—for example, 75 percent—and therefore liquidity figures prominently in portfolio allocation decisions.

VI Bubbles, Noise, and the Trading Process in Speculative Markets

Remarkably little is understood about the short-run behavior of prices and trading volumes in highly liquid asset markets such as stock and foreign exchange markets. These markets, at times, experience very sharp price movements while at the same time other variables that are presumed to be economic fundamentals change very little, if at all. The surprisingly large value of trading in some markets also is difficult to explain based on the ideas in the economics literature about the fundamental reasons for trading. Although the reasons for short-term movements in asset prices remain enigmatic, the macroeconomic fundamentals suggested by many traditional models explain asset prices reasonably well over longer periods of time. One possible reason for this contrast in model performance is the presence of unmodeled short-run market influences that tend to abate over longer periods. While the long-run success of models is somewhat comforting, it is of little help to policymakers or short-term investors, who are under considerable pressure to understand short-term market movements.

As a first step toward understanding short-term developments in asset markets, researchers have introduced various pathologies into economic models. These pathologies, such as rational asset-price bubbles, are persuasive rationalizations only to the extent that they place restrictions on data. In other words, introducing pathologies as an explanation of asset-market behavior is useful only if the proposed pathologies are inconsistent with some types of behavior. An explanation that is consistent with everything is no explanation at all. In this section we introduce some of the pathologies, develop some of the restrictions they impose on the data, and review some relevant empirical work.

Problems Explaining Short-Run Price Movements

A simple and appealing model of rational stock pricing implies that the value of a share is related to the expected stream of dividends paid by that share in the future (the discounted dividend model).95 One influential extension of this model found that while prices deviate from the predictions of the basic model in the short run,96 they have a tendency to move back toward the model’s predictions in the long run; that is, the model works much better over a long adjustment period than it does over a short one.97 The properly of “mean reversion” is strongly supported by rigorous testing procedures. A similar study of the foreign exchange market found that while macro-economic models explained exchange rates reasonably well in the long run, the forecasts of these models were substantially worse in the short run than the naive prediction that the exchange rate would stay constant.98

As an illustration of this tendency toward long-term mean reversion, consider the simple relative purchasing power parity (PPP) model of exchange rates. This model postulates that the percent change in the exchange rate between the currencies of two countries is equal to the inflation differential between those countries. Chart II.1 plots pairs of annual percent changes in dollar exchange rates for the industrial countries against each country’s annual inflation rate relative to that of the United States.99 If the PPP hypothesis is correct, then these pairs should coincide with the 45-degree diagonal line. On a year-to-year basis, the inflation differential is not closely related to exchange rate changes (see Panel 1 of Chart II.1). For countries with low-inflation differentials—10 percent or less in absolute value—the inflation differential is uninformative about the rate of depreciation of the exchange rate. The remaining panels in the chart plot the data averaged over 5, 10, and 20 years. Over long periods of time, inflation differentials are quite helpful in understanding average exchange rate changes (see Panels 2 through 4 of Chart II.1), and there does appear to be a strong tendency for exchange rate changes and inflation differentials to coincide.100

Chart II.1.Relative PPP in IFS Industrial Countries, 1972–941

(In U.S. dollar terms)

Source: International Monetary Fund, International Financial Statistics (IFS).

1 Excluding San Marino from the countries listed as industrial countries in the IFS.

2Five-year averages are for 1972–76, 1977–81, 1982–86, and 1987–94.

3Ten-year averages are for 1972–81 and 1982–94.

The failure of economic models to explain short-run movements in asset prices has generated interest in both academic and policymaking settings. Indeed, to many observers, the most glaring weakness of economic models is their inability to predict and to explain—after the fact—periods of turbulence in financial markets. There are several recent examples of sharp price and volume movements that models could neither predict ex ante nor describe ex post. During the October 19, 1987 stock market “break,” the Dow Jones Industrial Average (DJIA) fell 22.6 percent in one day, and totaled a 31 percent drop over four trading days starting on October 13, 1987.101 During the subsequent October 13, 1989 mini-break the DJIA declined almost 7 percent. Other examples include the attack on currency parities in the European Economic and Monetary Union (EMU) in 1992–93, the rise and fall of the “bubble economy” in Japan in the early 1990s, the fall of the Mexican peso during and after December 1994 and the coincident stock market declines in many developing countries. Examples such as these seem particularly relevant to newly established stock markets all over the world and help place in context experiences such as the MMM stock scheme in Russia during 1994.

In each of these examples and in many other cases, price movements appear to be independent of movements in traditional market fundamentals. Researchers have only recently documented and accepted these short-run problems with popular models. Formal mathematical treatment of these problems is more recent still. The next section is very much a “report from the front” on the models that researchers are attempting to construct and test in order to account for short-run asset price movements.

A New Generation of Asset-Market Models

A new generation of financial models has been developed to try to overcome some of the weaknesses of the traditional models and to help explain some of the more dramatic market events. To explain short-term movements in prices, the new work differentiates between classes of investors, whereas traditional models assumed that all investors were alike. These new models add microeconomic detail in such a way that investor diversity is intrinsic and essential. Models that differentiate between economic agents generally yield the same long-term qualitative predictions as models with homogeneous agents, however.

Asset-Market Bubbles

Although a precise definition of a bubble requires a model of a particular market, when an asset price rises solely because agents expect it to rise, the asset is said to be “on a bubble.” Two types of bubbles are studied in the literature: rational bubbles and other bubbles. Models of rational bubbles are interesting because the assumption of rationality produces simple statistical tests for the existence of bubbles. However, the assumption of rationality may be so restrictive that, in some cases, models do not explain the relevant data very well; that is, in some cases the model is rejected.

The rational bubble, which sounds like an oxymoron, is the mathematical formalization of the well-known greater fool theory. This theory postulates that an asset price is reasonable as long as a greater fool will later pay a sufficiently higher price for the asset. The value of a bubble-infected asset is driven, in part, by the expectation that the price of the asset will continue to increase. Once a rational bubble has started, investors do not, on average, expect to get rich from buying into the bubble. Some investors might reap high returns before the bubble bursts, but others, of (mathematical) necessity, will hold on to the asset too long and suffer a considerable loss during the inevitable crash.

A Model of a Stock Market Bubble

It is useful to illustrate precisely what is meant by a rational bubble in the context of a stock market model.102 Recent theoretical work indicates that bubbles may appear in models where generations of asset holders retire and are replaced by a new and sufficiently more wealthy generation of investors (the greater fools). This new generation is replaced by another one after a time, and so on. The appearance of successively more wealthy new generations of investors allows price bubbles to persist.103

Rational bubbles can occur in models in which the current price depends on the anticipated future price. One well-known, stock market model that generates bubbles is

Pt is the price of a share of stock now, D¯ is the constant dividend paid to holders of the stock, Pt+1 and is the prospective sale price of the stock next period, ρ is the discount rate (normally about .95 for inflation-adjusted annual data). This model adopts the fiction that all future dividends are constant and known currently and future prices are known also. Even with such simplifications, equation (1) is still a single equation in two unknowns, Pt and Pt+1. One equation in two unknowns determines a line relating Pt to Pt+1, not a point determining Pt. This simple notion is the basis of all work on rational bubbles.

Because Pt and Pt+1 occur at different points in time the model’s solution, multiplicity has an intertemporal dimension. According to equation (1), any price set this period will set next period’s price also; and setting next period’s price will determine price two periods into the future, and so on. This process is illustrated in Chart II.2, where equation (1) appears as the dashed line AA. Any point on AA is a solution to equation (1). The diagonal line shows points where price is constant over time (Pt = Pt+1). The solution of equation (1) with constant price occurs at the intersection of the two lines, Pt=P¯=(ρ/(1ρ))·D¯ This is the fundamental solution of the model for Pt.104

Chart II.2.Stock-Market Bubbles

If price is different from P¯, equation (1) sets in motion some price dynamics. The arrows show the direction of price movement. If Pt is above P¯, Pt+1 must be even higher than Pt; if Pt is below, P¯t+1 must be even lower than Pt.105 Prices like Pt1 and Pt2, which are away from the fundamental price P¯, involve self-fulfilling expectations of future capital gains or losses. These departures from fundamentals are known as rational price bubbles.

Economists know very little about bubbles in practice, but quite a lot in theory. As an example, consider the price Pt1 in the Chart II.2. Read Pt1 off the horizontal axis and Pt+11 off the vertical axis. Evidently, since line AA is steeper than the diagonal (where Pt+1 equals Pt), Pt+11 is above Pt1. Of course, next period the higher Pt+11 will require an even higher capital gains, and so on. Current capital gains end up being justified by even larger future capital gains. For prices above the fundamental solution the market would take off on a self-fulfilling positive bubble as indicated by the arrow going northeast in Chart II.2.

Now suppose market price is below the fundamentals price (e.g.,Pt2 < P¯). This may appear to result in a negative bubble, but a bit of introspection reveals that negative bubbles cannot exist. Following the negative bubble’s path to the southwest, it is clear that the price must eventually fall below zero.106 This makes no sense for assets like stocks, which have limited liability—shareholders can lose no more than their initial investment. Any limited liability shareholder would gladly renounce ownership rather than hold an “asset” with negative value. A negative bubble, therefore, cannot be sustained by a rational forward-looking market.

Positive rational bubbles, on the other hand, make sense in some models. Their existence and description, therefore, becomes an empirical question. To study actual data, the simplifications of constant dividends and perfect anticipation of future prices and dividends must be removed. These simplifications are normally replaced by the condition that investors forecast future prices, dividends, and any other relevant variables as efficiently as possible, using all available information appropriately. This assumption, known as “rational expectations,” preserves the mathematical structure so far developed, but allows investors to be wrong—perhaps disastrously so.107

Testing for Rational Bubbles

Models with rational bubbles have been popular in the literature because they embody the idea that bubbles drive prices away from their fundamental values. These models are also popular because they produce sharply defined predictions and hypotheses about prices and thus are amenable to empirical testing. The prospective bubbles cannot, however, be effectively separated from the models that generate them. Every test for bubbles is in the context of a model and is part of a joint hypothesis that the model is correct, that expectations are formed rationally, and that bubbles are absent. If the investigator is willing to maintain that the model is correct and that expectations are formed rationally and the hypothesis is statistically rejected, it may be concluded that bubbles may not be absent.

Models like that described in equation (1) generate testable hypotheses about market prices. The bubbles tests involve the difference between actual prices and a counterfactual price series—the one generated by the model and based on actual model-suggested fundamentals. The tests then use statistical theory to judge whether the difference is larger than would be given by chance and has the properties suggested by the relevant bubble.108 For example, in the model portrayed in equation (1) the market fundamentals price is the constant Pt=(ρ/(1ρ))·D¯. In a bubbles test based on equation (1), if actual prices were significantly different from this constant and were growing on average at the rate (1/ρ) then the investigator would not reject the presence of a rational bubble in this market.

Existing research constructs the counterfactual series in two ways. First, mathematical models are used, as above, to construct price series that would have been based appropriately on fundamentals.109 The second approach compares prices in markets that are thought not to contain bubbles with prices in similar markets where bubbles are suspected.110 Tests for bubbles always involve these counterfactual comparisons; in the first case, against the implications of a mathematical model and in the second case, against a similar market where bubbles are thought to be absent.

The tests are only as good as their counterfactuals, however. If one has great confidence in a model and the actual price series is different from the model-generated counterfactual in ways predicted by the presence of bubbles, then one might conclude that the data are consistent with bubbles.111 If confidence in the model is low, however, then an apparent bubbles finding might be attributed to misspecification of the model-generated counterfactual series. For example, adopting equation (1) as the stock market counterfactual, one would generate a constant “fundamental” price in a market in which dividends actually grow on average but in an unpredictable way. Maintaining the model to be correct would lead to an apparent finding of a stock market bubble. In the example, bubbles might indeed be present, but because the model is so clearly wrong, a conclusion that bubbles are present would be ill founded because the counterfactual model is so weak.

Historical Bubble Reports: Tulipmania

Two kinds of empirical work have investigated bubbles: studies of purported historical bubbles, such as the Dutch Tulipmania of the 1630s, and more modern studies of prices and fundamentals in, for example, stock markets, foreign exchange markets, and land markets. Descriptions of events in the Dutch market for tulips in the period 1634–37 are often included in historical accounts of speculative market excesses.112 During 1634–37, the Semper Augustus tulip bulb sold for 5500 guilders, a gold-price equivalent to $50,000 at $450 per ounce.113 Historical descriptions of these events indicate that a speculative frenzy overtook the bulb market in early 1637 and drove prices to extreme heights, which was followed by a crash. After the crash, bulbs could not fetch prices that were as high as 10 percent of peak prices.

A more recent study using data from this period found that speculative run-ups and crashes were typical patterns in the historical flower bulb market.114 Indeed, the pattern still persists in some ways; a 1987 prototype lily bulb sold for $480,000. The study compares the path of tulip bulb prices over 1634–37, for which bubbles are suspected, to bulb price paths in other markets that are thought not to contain bubbles. As the difference was found to be not significant, the so-called Dutch Tulip bubble probably was not a rational bubble.

Once the study’s testing methodology is accepted, the explanation that Tulipmania is driven by the same fundamentals that drives other flower prices is convincing for some of the data. Not all of the events of 1637 can be rationalized, however. During the winter of that year, futures markets in bulbs flourished in taverns. In this market, the prices for common bulbs rose and exploded in a pattern otherwise unique to the rarest species.

Bubbles have also been studied using formal econometric models.115 The first such test examined whether the tremendous acceleration of inflation during 1922–23 in post–World War I Germany was partly due to a price level bubble. The hypothesis that no bubbles were present was not rejected. Possible price level bubbles in one economy, of course, have exchange rate implications. Other related work testing for foreign exchange market bubbles was unsupportive of the existence of such bubbles.116

Tests for bubbles also were applied to stock prices.117 Studies have found evidence for bubbles in the stock market using state-of-the-art methods. The tests, however, suffer from the problem that a bubble can be identified only relative to a particular model of price fundamentals. A fully credible model has remained elusive in the stock market. Rejecting the hypothesis that bubbles are absent could just as well indicate that the model generating the counterfactuals is wrong.

The Japanese Bubble Economy

Among the best-known modern asset-market disturbances is the “bubble economy” in Japan in the late 1980s. Stock prices and land prices in Japan increased at a very rapid pace.118 The Nikkei 225 index of stock prices went from 9,500 in December 1983 to 18,700 in December 1986, and then to 38,000 in 1989, more than doubling every three years. Average land prices in Tokyo doubled between the end of 1986 and the end of 1988. Although the timing of price increases differed across different types of assets, the capital gains from owning stocks and real estate greatly exceeded the interest rates. High returns on property attracted many investors, driving prices up further, which is characteristic of a bubble process.

Stock prices peaked at the end of 1989, and land prices peaked around 1990–91. Once a turning point was reached, asset prices plummeted. In the process, stock and land prices lost most of their gains from the rise during the bubble period. Asset owners suffered capital losses; many real estate companies and developers, who had borrowed to acquire assets, became insolvent. Although there is some work testing whether the price movements in Japan can be explained by microeconomic or macroeconomic fundamentals, the evidence appears to be mixed.119

Other Bubbles

Some researchers feel that the assumption of rationality is too restrictive. Although the possibility that asset prices in models could be driven by expectations rather than by fundamentals has been in the literature for some time, the older literature was unsatisfactory in that not only were destabilizing expectations not self-fulfilling, they were increasingly self-deluding.120

Recent work in this tradition does not require rationality and need not be self-deluding in the long run. This work examines a foreign exchange market populated by three types of traders: fundamentalists, chartists, and portfolio managers.121 Fundamentalists assume that the exchange rate will take on its fundamental value; chartists forecast the future exchange rate by extrapolating from past exchange rates; and portfolio managers forecast an exchange rate that is a weighted average of those expected by the chartists and fundamentalists. The type of adaptive learning used by the model’s portfolio managers generates nonlinear dynamics. For some specifications of this model, price moves away from the path that would prevail if all expectations were formed by the fundamentalists, due to nonrationality. For different parameter configurations, the model is able to mimic the speculative run-ups and crashes that characterize foreign exchange markets.

Another notion of a bubble has been investigated in experiments with a computerized stock market, allowing subjects to “trade” in the market.122 In the experiments, both researchers and experimental subjects (college students and businessmen) knew the price that should be determined by fundamentals, the random final-period cashflow. One would expect that in such a controlled setting, the stock price would quickly attain the expected value of the final payout, but it did not. In 14 of 22 experiments, the authors found a boom-bust pattern typical of a bubble. This kind of bubble is different from those discussed previously because the subjects knew all possible terminal prices, but prices still fluctuated wildly. The authors interpreted this as the result of traders’ learning about the market and possible strategic behavior by market participants.

Noise Trading

Much empirical work suggests that trading volume is related to price changes. This leads some to suspect that something in the trading process is generating what appears to be excess price volatility. Several new approaches have emerged to address this possibility.

Trading based only on rationally processed differences in information is impossible. If traders agree on what information is important, and how to interpret that information, differences in information will generate no trading. Suppose there are two traders: one who knows the value of the dividend to be paid out next period, and one who does not. Suppose also that the information is favorable; the stock is a bargain at current prices, given knowledge of the future dividend. The informed trader will want to profit from his information, and buy the uninformed traders’ shares. When the informed trader offers to pay more for the shares of the second than the going price, he signals to the uninformed trader that the prospective buyer has superior information, and that his information indicates that the market price is too low. The uninformed trader will rationally refuse the offer, now knowing that the market price is too low. In fact, he will refuse any price that the informed trader finds acceptable (up to the actual value of the stock given dividend information). Hence, no trading will take place. This simple but powerful result is known as a no-trade theorem and prompted several responses.123

In one type of response non-information-based reasons are introduced for trade. Some people sell stocks to finance purchases, adjust portfolios, generate retirement income, and so on. In these new models, some traders are (realistically) allowed to have fundamentally changing life circumstances—death, retirement, changes in tastes, or changes in labor income, all of which would prompt rational trade. Likewise, some people no doubt trade because they do not understand the market, because they receive bad investment advice, because they think they can outguess other traders, or simply because they like to trade. Hence, in another type of response, the full rationality requirement is relaxed so that some traders do not have a complete understanding of the information-processing abilities of other traders or, more simply, about how to process information themselves.

These latter traders, who do not act rationally, are known as noise traders. The idea that traders act in a fully rational manner, whose expectations are consistent with the reality of the model, is a modern innovation.124 Consideration of viable noise traders in addition to rational traders is modern also and is different from the idea of rational bubbles. Roughly speaking, noise trading involves short-run deviations from a fundamentals equilibrium, while bubbles involve deviations that are expected to last a long time.125

Noise trading is not necessarily all bad. A well-known contribution in this area126 pointed out that in the absence of frequent trading, markets are illiquid—death, retirement, and other circumstances simply cannot generate much trading. Illiquid markets do a poor job of impounding relevant information into prices; the liquidity generated by noise traders makes markets more efficient in revealing this information.

While the early work on noise was quite general, later work became more specific by concentrating on the price effects of noise.127 In this work, there are two types of traders. The first kind is a typical utility-maximizing “information trader” who knows about fundamentals and takes informed positions. The second kind is a “noise trader,” who takes positions that are not in accordance with standard economic principles. Either the noise trader is behaving irrationally for unspecified reasons, escapes the norms of economists’ models, or simply likes to trade.

At one time, most mainstream economists discounted the notion of noise traders because traders that are systematically wrong would quickly lose all their money to fundamentals traders, and be driven out of the market.128 However, this new model of noise traders shows that this need not happen. Noise traders’ activity creates price fluctuations in risky assets they hold; this makes fundamentalists, who dislike risk more than noise traders do, less apt to hold these assets. The noise traders thus create a space for themselves in the market, which is insulated from the information traders by the very risk the noise traders create.

Herd Behavior

In most of the work discussed so far, it has been assumed that investors make their decisions ignorant of the decisions of others. A different kind of assumption is explored in models of herd behavior.129 In this work, not all information is publicly available, and agents make decisions in turn. The sequencing of decisions is crucial since decision-makers toward the end of the queue can learn from the decisions made by those before them. Hence, individual decisions can have externalities, and the cumulative effect of these externalities may lead to inefficient investment allocations. With this kind of information, an investor’s decision is predicated both on his private information about market fundamentals and by his observations of others’ decisions. The decision of other investors yields information, as it might reveal what other investors know.

As an example, suppose that all investors have access to public information about an investment project, such as balance sheets and accountants reports. Suppose also that each investor has a private information source that may or may not be reliable. Suppose there are two investors, A and B, considering one project, and investor A is first in line. Suppose investor A has a “tip” that the project is good, while B has a tip that it is bad. A, going first, follows his tip and invests. B observes A’s decision, and weighing his own information decides to follow A and invest, inferring that A must know something he doesn’t. Now suppose that, in reality, A’s private information was wrong. Although B may have correct private information, that private signal was swamped by knowledge of As actions. The asset market would misprice the project because of the externality imposed by sequential decision making with private information.130 Models based on similar ideas have been used to explain diverse phenomena, such as political bandwagons; medical fads known as treatment-caused epidemics (iatroepidemics), such as tonsillectomies; and financial-market behavior, such as bank runs in which depositors’ observations of other depositors actions can trigger a cascade of further withdrawals.

Conclusion

At one time it was widely asserted that pathologies such as bubbles and noise were inconsistent with well-functioning rational markets. Careful theoretical work, however, has established conditions such that these alleged pathologies could logically endure in models. This, no small achievement, has altered the way many researchers approach markets.

Establishing the pathologies in theory is quite different from confirming that they are relevant to actual markets. The biggest obstacle is, again, that the tests are only as good as the counterfactuals. This applies to tests for bubbles, noise, or any other source of concern about market outcomes. All of the tests require a comparison of actual market outcomes against a model of the outcomes the market would have generated had the alleged pathologies been absent.

The current generation of models of asset pricing based on fundamentals, the counterfactuals, is simply not very convincing. According to these models virtually all asset markets are too volatile and always have been so. While that is a consistent position, it is not persuasive or informative as to the urgency of concerns about asset-market volatility.

When researchers see prices change in ways that are inconsistent with their models, they have two polar responses. One response is that the model is wrong, and the market, right. The solution is to change the model. The other response is that the model is wrong because certain pathologies inhabit the market; the model is right, and the market, wrong. Although respectable researchers inhabit both poles, most are somewhere in the middle, concerned about large and unmotivated price changes but unimpressed by theories that have not held up to empirical testing.

Bibliography

    “America’s Top 300 Money Managers”Institutional Investor (July1994) pp. 11348.

    BalassaBelaThe Purchasing Power Parity Doctrine: A ReappraisalJournal of Political Economy Vol. 72 No. 6 (December1964) pp. 58496.

    Bank for International SettlementsPublic Disclosure of Market and Credit Risks by Financial Intermediaries” (Basle: Bank for International SettlementsSeptember1994).

    Bank for International SettlementsIssues of Measurement Related to Market Size and Macroprudential Risks in Derivatives Markets” (Basle: Bank for International SettlementsFebruary1995).

    Basle Committee on Banking SupervisionInternational Convergence of Capital Measurement and Capital Standards” (Basle: Basle Committee on Banking SupervisionJuly1988).

    Basle Committee on Banking Supervision (1993a) “The Prudential Supervision of Netting, Market Risks and Interest Rate Risk” (Basle: Basle Committee on Banking SupervisionApril1993).

    Basle Committee on Banking Supervision (1993b) “The Supervisory Recognition of Netting for Capital Adequacy Purposes” (Basle: Basle Committee on Banking SupervisionApril1993).

    Basle Committee on Banking Supervision (1993c) “The Supervisory Treatment of Market Risks” (Basle: Basle Committee on Banking SupervisionApril1993).

    Basle Committee on Banking Supervision (1993d) “Measurement of Banks’ Exposure to Interest Rate Risk” (Basle: Basle Committee on Banking SupervisionApril1993).

    Basle Committee on Banking Supervision (1994a) “Basle Capital Accord: The Treatment of the Credit Risk Associated with Certain Off-Balance-Sheet Items” (Basle: Basle Committee on Banking SupervisionJuly1994).

    Basle Committee on Banking Supervision (1994b) “Risk Management Guidelines for Derivatives” (Basle: Basle Committee on Banking SupervisionJuly1994).

    Basle Committee on Banking Supervision (1994c) “Report of International Developments in Banking Supervision,” Report No. 9 (Basle: Basle Committee on Banking SupervisionSeptember1994).

    Basle Committee on Banking Supervision (1995a) “Proposal to Issue a Supplement to the Basle Capital Accord to Cover Market Risks” (Basle: Basle Committee on Banking SupervisionMarch1995).

    Basle Committee on Banking Supervision (1995b) “An Internal Model-Based Approach to Market Risk Capital Requirements” (Basle: Basle Committee on Banking SupervisionMarch1995).

    Basle Committee on Banking Supervision (1995c) “Planned Supplement to the Capital Accord to Incorporate Market Risks” (Basle: Basle Committee on Banking SupervisionMarch1995).

    BennettRosemary and DavidShirreffLet’s Bash the Hedge FundsEuromoney (April1994) pp. 2633.

    BikchandaniSushilDavidHirschleifer and IvoWelchA Theory of Fads, Fashion, Custom and Cultural Change as Informational CascadesJournal of Political Economy Vol. 100 (October1992) pp. 9921025.

    BisignanoJosephThe Internationalization of Financial Markets: Measurement, Benefits, and Unexpected InterdependenceCahiers économiques et monétaires no. 43 Banque de France (November1994) pp. 971.

    BlackFischerNoiseJournal of Finance Vol. 41 (July1986) pp. 52943.

    BlanchardOliver and MarkWatsonBubbles, Rational Expectations and Financial Markets” in Crises in the Economic and Financial Structureed. by P.Wachtel (Lexington: Lexington Books1982).

    BodieZviShortfall Risk and Pension Fund ManagementFinancial Analysts Journal (May-June1991) pp. 5761.

    “Books of Revelations?”Risk Vol. 7 (September1994) pp. 91102.

    BorchardtKlaus-DieterThe ABC of Community Law” (Luxembourg: Office of Official Publications of the European Communities1993).

    CaganPhillipThe Monetary Dynamics of Hyperinflation” in Studies in the Quantity Theory of Moneyed. by MiltonFriedman (Chicago: University of Chicago Press1956) pp. 25117.

    ChuhanPunamAre Institutional Investors an Important Source of Portfolio Investment in Emerging Markets?World Bank Policy Research Working Paper No. 1243 (Washington: The World Bank1994).

    ClaessensStijn and SudarshanGooptueds.Portfolio Investment in Developing Countries World Bank Discussion Papers 228 (Washington: The World Bank1993).

    DavisE.P.International Diversification of Institutional InvestorsBank of England Discussion Paper No. 44 (September1991) pp. 153.

    DeLongJ.Bradford and othersThe Size and Incidence of the Losses from Noise TradingThe Journal of Finance Vol. 44 (July1989) pp. 68195.

    “Emerging Markets Debt Comes of Age”Institutional Investor (April1994) pp. 6368.

    European Community Council Directive 93/6/EECOn the Capital Adequacy of Investments Firms and Credit InstitutionsOfficial Journal of the European Communities (LuxembourgMarch1993).

    “Fall Guys?”Business WeekApril251994 pp. 11621.

    FamaEugene and KennethFrenchPermanent and Temporary Components of Stock PricesJournal of Political Economy Vol. 96 (April1988) pp. 24673.

    Federal Reserve Bank of New YorkWholesale Transaction Code of Conduct” (New York: Federal Reserve Bank of New YorkJanuary1995).

    Financial Accounting Standards BoardStatement of Financial Accounting Standards No. 119: Disclosure About Derivative Financial Instruments and Fair Value of Financial Instruments” (Norwalk, Connecticut: Financial Accounting Standards BoardOctober1994).

    FloodRobert and PeterGarberMarket Fundamentals Versus Price Level BubblesJournal of Political Economy Vol. 88 (August1980) pp. 74570.

    FloodRobert and RobertHodrickOn Testing for Speculative BubblesJournal of Economic Perspectives Vol. 4 (Spring1990) pp. 85101.

    FrankelJeffrey A. and Kenneth A.FrootChartists, Fundamentalists, and Trading in the Foreign Exchange MarketAmerican Economic Review Papers and Proceedings Vol. 80 (May1990) pp. 18185.

    Fraser InstituteSurvey of Senior Investment Managers: Results for Winter 1994/95 (Vancouver: December191994).

    FrenchKenneth R. and James M.PoterbaInvestor Diversification and International Equity MarketsAmerican Economic Review Papers and Proceedings Vol. 81 (May1991) pp. 22226.

    FriedmanMiltonThe Case for Flexible Exchange Rates” in Essays in Positive Economicsed. by MiltonFriedman (Chicago: University of Chicago Press1953).

    FrootKenneth and MauriceObstfeldIntrinsic Bubbles: The Case of Stock PricesAmencan Economic Review Vol. 81 (December1991) pp. 11891214.

    “Fund Management”Special Supplement International Financial Law Review (February1994).

    GarberPeterTulipmaniaJournal of Political Economy Vol. 97 (June1989) pp. 53560.

    GarberPeterFamous First BubblesJournal of Economic Perspectives Vol. 4 (Spring1990) pp. 3554.

    GooptuSudarshanPortfolio Investment Flows to Emerging Markets” in Portfolio Investment in Developing Countriesed. by StijnClaessens and SudarshanGooptu World Bank Discussion Papers 228 (Washington: The World Bank1993) pp. 4577.

    GrossmanSanford and JosephStiglitzOn the Impossibility of Informationally Efficient MarketsAmerican Economic Review Vol. 70 (June1980) pp. 393408.

    Group of ThirtyDerivatives: Practices and Principles (Washington: Group of ThirtyJuly1993).

    Group of Thirty (1994a) 1994 Annual Report (Washington: Group of Thirty1994).

    Group of Thirty (1994b) “Derivatives: Practices and Principles—Follow-Up Surveys of Industry Practice” (Washington: Group of ThirtyDecember1994).

    HayworthPCBasle Committee on Bank Supervision” in The New Palgrave Dictionary of Money and Finance Vol. 1ed. by PeterNewmanMurrayMilgate and JohnEatwell (New York: Stockton Press1992) pp. 18587.

    HoffmaisterAlexander W. and Garry J.SchinasiAsset Prices, Financial Liberalization, and Inflation in Japan” in Saving Behavior and the Asset Price “Bubble” in Japan: Analytical Issuesed. by UlrichBaum-gartner and GuyMeredith IMF Occasional Paper No. 124 (Washington: International Monetary FundApril1995).

    HowellMichael J.Institutional Investors and Emerging Stock Markets” in Portfolio Investment in Developing Countriesed. by StijnClaessens and SudarshanGooptu World Bank Discussion Papers 228 (Washington: The World Bank1993) pp. 7887.

    Institute of International Finance Inc.A Preliminary Framework for Public Disclosure of Derivatives Activities and Related Credit Exposures” (Washington: Institute of International Finance, Inc.August1994).

    International Finance CorporationEmerging Stock Markets Fact Book 1994 (Washington: International Finance Corporation1994).

    International Monetary FundInternational Capital Markets: Part I. Exchange Rate Management and International Capital Flows World Economic and Financial Surveys (Washington: International Monetary FundApril1993).

    Group of ThirtyInternational Financial Statistics (Washington: International Monetary FundAugust1993).

    Group of ThirtyInternational Capital Markets: Part II. Developments Prospects and Policy Issues World Economic and Financial Surveys (Washington: International Monetary FundSeptember1994).

    International Organization of Securities Commissions1994 Annual Report (Montreal: International Organization of Securities Commissions1994).

    ItoTakatoshiThe Japanese Economy (Cambridge, Massachusetts: MIT Press1992).

    ItoTakatoshiThe Land/Housing Problem in JapanThe Journal of the Japanese and International Economies Vol. 7 (March1993) pp. 131.

    ItoTakatoshi and TokuoIwaisakoExplaining Asset Bubbles in JapanBank of Japan Monetary and Economic Studies (forthcoming).

    ItoTakatoshi and Keiko HironoNosseEfficiency of the Tokyo Housing MarketBank of Japan Monetary and Economic Studies Vol. 11 (July1993) pp. 132.

    J.P. Morgan & Co Incorporated (1994a) “Introduction to RiskMetrics” (New York: Morgan Guaranty Trust CompanyOctober1994).

    J.P. Morgan & Co Incorporated (1994b) “RiskMetrics Technical Document” (New York: Morgan Guaranty Trust CompanyOctober1994).

    KupiecPaul and JamesO’BrienA Pre-Commitment Approach to Capital Requirements for Market Risk” (mimeograph Washington: Board of Governors of the Federal Reserve SystemMarch1995).

    KyleAlbert S.Informed Speculation with Imperfect CompetitionReview of Economic Studies Vol. 56 (July1989) pp. 31756.

    LaWareJohn P.Testimony on Hedge Funds Before the Committee on Banking Finance and Urban Affairs of the U.S. House of Representatives (Washington: April131994).

    McKayCharlesMemoirs of Popular Delusions and the Extraordinary Madness of Crowds (London: Office Nar. Illustrated Library 2nd ed.1852).

    MeeseRichard and KennethRogoffOut-of-Sample Failure of Empirical Exchange Rate Models: Sampling Error of Misspecification?” in Exchange Rates and International Macroeconomicsed. by Jacob A.Frenkel (Chicago: University of Chicago Press1983).

    MilgromPaul and NancyStokeyInformation, Trade and Common KnowledgeJournal of Economic Theory Vol. 26 (January1982) pp. 1727.

    Moody’s Investors ServiceAsset Quality Looms as Major Problem for Mexican BanksJanuary1995.

    Money Market DirectoriesDirectory of Pension Funds and their Investment Managers 1995 (McGraw-Hill 25th ed.1995).

    MuthJohnRational Expectations and the Theory of Price MovementsEconometrica Vol. 29 (July1961) pp. 31535.

    NoelEmileWorking Together—The Institutions of the European Community” (Luxembourg: Office of Official Publications of the European Communities1993).

    Organization for Economic Cooperation and DevelopmentFinancial Statistics Monthly various issues.

    PaganoMarco (1989a) “Trading Volume and Asset LiquidityQuarterly Journal of Economics Vol. 104 (May1989) pp. 25574.

    PaganoMarco (1989b) “Endogenous Market Thinness and Stock Price VolatilityReview of Economic Studies Vol. 56 (April1989) pp. 26987.

    RichardsonPortiaOutlook 95: The Rugged Singles SceneInstitutional Investor (January1995) pp. 11525.

    SchinasiGarry J. and MonicaHargraves‘Boom and Bust’ in Asset Markets in the 1980s: Causes and Consequences” in Staff Studies for the World Economic Outlook World Economic and Financial Surveys (Washington: International Monetary Fund1993) pp. 127.

    ShillerRobertDo Stock Prices Move Too Much to Be Justified by Subsequent Changes in Dividends?American Economic Review Vol. 71 (June1981) pp. 42135.

    ShillerRobertMarket Volatility (Boston: MIT Press1991).

    SmithVernon L.GerrySuchanek and Arlington W.WilliamsBubbles, Crashes and Endogenous Expectations in Experimental Spot Asset MarketsEconometrica Vol. 56 (September1988) pp. 111951.

    Technical Committee of the International Organization of Securities CommissionsOperational and Financial Risk Management Control Mechanisms for Over-the-Counter Derivatives Activities of Regulated Securities Firms” (Montreal: International Organization of Securities CommissionsJuly1994).

    TiroleJeanAsset Bubbles and Overlapping GenerationsEconometrica Vol. 53 (September1985) pp. 1071100.

    UedaKazuoAre Japanese Stock Prices Too High?Journal of the Japanese and International Economies Vol. 4 (December1990) pp. 35170.

    United States General Accounting OfficeFinancial Derivatives: Actions Needed to Protect the Financial System” (Washington: U.S. General Accounting OfficeMay1994).

    WestKennethBubbles, Fads and Stock Price Volatility Tests: A Partial EvaluationJournal of Finance Vol. 43 (July1988) pp. 63956.

Banks in general face the following risk elements: credit risk—that a counterparty might default on its position; market risk—a loss due to unexpected general market price and interest rate changes; operation risk—a loss due to human error, fraud, or the lack of internal controls; legal risk—related to the legal status of a contract; liquidity risk—that a position cannot be sold very quickly without a huge price concession; settlement risk—the market and credit risk exposure during the settlement period; and specific risk—a decline in value of a particular position that is not the result of a general market movement.

The 1988 Basle Accord establishes a capital requirement as a percentage of the credit risk exposure of a bank. Under this approach, each on-balance-sheet item is assigned a risk weight reflecting the credit quality of the counterparty involved, which is determined by such factors as whether the counterparty is a government or a private institution, and whether the counterparty is from an OECD country or not. Off-balance-sheet interest rate and exchange rate contracts are first converted to a so-called credit-equivalent amount under one of two permitted conversion methods. The risk weights are then applied to these credit equivalent amounts as if they are on-balance-sheet positions. The two conversion methods are (1) the so-called current exposure method, under which the credit equivalent is calculated as the current replacement cost of the position plus an add-on factor equal to a percentage of the notional principal; (2) the so-called original exposure method, under which the credit equivalent is calculated by multiplying the notional principal by a conversion factor.

A detailed discussion of the Basle Proposal and the CAD is provided in International Monetary Fund (1994).

See Basle Committee on Banking Supervision (1988) and (1994a). The collapse of Barings suggests that market risk, operational risk, and internal management controls are very important elements of risk management. At the end of 1993, Barings had a total capital ratio well in excess of the 8 percent requirement, and in January 1995, it was still considered to be a safe bank. By the end of March 1995, however, Barings was in administration. The failure of a well-capitalized institution raises important questions about the adequacy of current regulatory capital requirements and the role they play in safeguarding individual financial institutions from financial distress.

See Basle Committee on Banking Supervision (1993ad).

Spread risk is the risk that the relationship between the prices or yields of two similar, but not identical, assets of the same maturity will change. Gap risk is the risk that the relationship between the prices or yields of two instruments of the same type, but of different maturities, will change.

Barings, which collapsed following a loss of about $1 billion from derivatives trading, was believed to have entered initially into a short-straddle position, before the Kobe earthquake, that took a bet on low volatility in the Japanese stock market.

A dynamic trading strategy is one in which the portfolio of assets is changed continuously according to a certain rule so as to achieve a risk-exposure target.

The Group of Thirty, in its report on derivatives released in July 1993, has highly recommended value-at-risk as a useful way to describe market risk exposure. The Fisher report, issued by the Bank for International Settlements (BIS) in September 1994, has also recommended banks to disclose their value-at-risk.

The asset normal approach is the method employed by J.P. Morgan’s highly publicized RiskMetrics, risk-management system. Under this distributional assumption, the standard deviation of portfolio return can be computed from the standard deviations of the asset returns, their correlations, and the individual weights of the assets in the portfolio using a standard statistical formula. The value-at-tisk under a 99 percent confidence interval is simply the level of loss corresponding to a 2.32 standard deviation drop in the value of the portfolio.

Because the value of an option relates to the price of the underlying asset in a nonlinear fashion, even if the underlying asset return is normally distributed, the return on an option position will not be normally distributed.

The delta of an option is the sensitivity of the option price with respect to changes in the value of the underlying asset.

The idea is to apply a first-order Taylor approximation to the nonlinear option payoff.

The gamma of an option is the sensitivity of the option’s delta to a change in the price of the underlying asset. The gamma risk is therefore the risk that the delta of an option might change.

The approach utilizes a second-order approximation instead of a first-order approximation.

Vega risk is the risk of a change in the value of an option due to a change in the volatility of the underlying asset.

The historical (bootstrapping) approach does not rely on any assumption about the statistical distribution of asset returns. Basically, it uses past data to construct a histogram that is then used as the distribution of asset return. Hypothetical price changes are drawn randomly and repeatedly from this probability distribution to obtain an estimate of the value-at-risk. An obvious advantage is that the results are not dependent on assumptions regarding the probability distribution of asset returns. A disadvantage is that the histogram from past data might not be a very accurate description of the asset return distribution in the future. Furthermore, the method is computationally very intensive and is therefore not very practical for a large portfolio.

The factor-push approach is quite different from the other approaches in that the focus is on the risk factors themselves. The idea is that each factor will be pushed toward a direction that reduces the value of the portfolio. A large loss value produced by large changes in the factors is taken to be the value-at-risk. This approach ignores the combined effects of the factors and the correlations among the factors.

The maximum loss approach computes the maximum portfolio loss under the constraint that the probability of the combination of price moves is not bigger than, say, 5 percent of the time. That is, this approach locates the worst scenario while making certain distributional assumptions. The approach is computationally very intensive.

In J.P. Morgan’s RiskMetrics system, an exponentially declining weight approach, which assigns a smaller weight to an older observation, is used in the estimation of variances and covariances.

Box II.1 gives examples of value-at-risk calculations for a portfolio using different methods.

In the April 1993 Basle Proposal, vertical disallowance is an adjustment made to the capital charge for offsetting long- and short-debt securities positions in the same time band. The proposal slots debt securities into 13 time bands according to their maturities in the process of calculating the capital charges. The vertical disallowance is for covering the gap risk related to long and short positions in the same time band that are similar but not identical securities and hence are not perfect hedges for one another.

See background paper “Mechanisms for International Cooperation in Regulation,” pp. 158–61, for a discussion of the mechanisms that currently exist for international coordination of financial regulation and supervision.

The collapse of Barings serves as an example of the importance of internal controls.

See Basle Committee on Banking Supervision (1994b) and Technical Committee of the International Organization of Securities Commissions (1994).

See the background paper “Regulatory Implications of the Barings Failure,” pp. 162–64.

End-users that have suffered big losses in derivatives or highly leveraged structured instruments include Orange County, Metallgesellschaft, Proctor & Gamble, and Gibson Greetings.

For instance, Barings lost close to $1 billion in less than two months on Nikkei index derivatives following the Kobe earthquake.

See Basle Committee on Banking Supervision (1988). Replacement value is the amount of money that would have to be paid to a third party to induce it to enter into a transaction to replace an existing position. Net replacement value of a portfolio is the replacement value of the portfolio after allowing long and short positions with the same counterparty to offset each other. An enforceable netting arrangement is an arrangement by which the offsetting of long and short positions with a counterparty is legally enforceable at the time of bankruptcy.

These rules include FAS No. 52, which applies only to foreign exchange transactions, not to currency options; FAS No. 80, which applies to futures but not forwards; FAS 105, which only requires the notional amount of OTC derivative positions be reported with their nature and terms; FAS No. 107, which requires the disclosure of fair value of related on- and off-balance-sheet derivatives; and, FASB Interpretation No. 39, which only allows netting of positions with a counterparty under a master netting arrangement.

The most publicized cases have been Proctor & Gamble and Gibson Greetings against Bankers Trust.

The Group of Thirty report emphasized the importance of qualitative disclosure including (1) management’s attitude to financial risk; (2) how instruments are used; (3) how risks are monitored and controlled; (4) accounting policies; (5) analysis of positions at the balance sheet date; and (6) analysis of credit risk inherent in those positions. See Group of Thirty (1993).

The Committee’s members are from Belgium, Canada, France, Germany, Italy, Japan, Luxembourg, the Netherlands, Sweden, Switzerland, the United Kingdom, and the United States.

Members of the EU are Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, the Netherlands, Portugal, Spain, Sweden, and the United Kingdom.

A description of the roles and current developments of the EU committees involved in banking supervision along with the other multinational banking supervisors organizations are in Basle Committee on Banking Supervision (1994c).

See Borchardt (1993) for more details.

The weighting of votes in the Commission is two votes each for France, Germany, Italy, Spain, and the United Kingdom; one vote each for the remaining seven member countries.

A qualified majority is 54 out of 76 votes. The weighting of votes in the Council is 10 votes each for France, Germany, Italy, and the United Kingdom; 8 votes for Spain; 5 votes each for Portugal, Greece, the Netherlands, and Belgium; 3 votes for Denmark and Ireland; and 2 votes for Luxembourg.

Records of the amount of time the issue was discussed at the EU Committee level before reaching the Commission are not available.

The EU also passes directives in the investment services area using the same procedures and a similar committee structure as described in the banking supervisory organization section above.

IOSCO (1994).

There are four regional committees that discuss specific problems of concern in their region. The regional groupings within IOSCO may be another reason why other regional securities firm supervisory organizations have not been formed.

If there is no government regulatory agency, a self-regulatory organization can become a regular member with voting rights.

When a country has multiple accountancy bodies, one is chosen to be its representative.

The Institute of International Finance, Inc., informational pamphlet.

“Administration” is a court-supervised reorganization. Barings plc is composed of Baring Brothers, the merchant banking operation, Baring Securities, the brokering and market-making operations, and Baring Asset Management, the fund-management operation; it also owns a 40 percent stake in Dillon Reed, the New York investment bank.

The Bank of England immediately and clearly indicated that the collapse of Barings was an isolated incident—with no systemic implications—and that the failure resulted from the activities of a single “rogue” trader who, “concealed this [the positions and losses] from his senior management, the management in London, from the local regulator and of course through them from ourselves, through collusion with his settlement people.” (See transcript of Governor George’s interview on “Today,” BBC Radio 4, February 27, 1995.)

More recent reports have suggested that Barings operations in London had sanctioned the overall trading strategy, and supplied funds to, the operation in Singapore.

To obtain an exemption from position limits, verification of a hedged position is required at periodic intervals and whenever there is a material change in the hedged position.

See “America’s Top 300 Money Managers” (1994).

These countries are Canada, Germany, Japan, the United Kingdom, and the United States.

The number of hedge funds located in Europe is not known, in large part because they are deliberately secretive.

LaWare (1994) and Bennett and Shirreff (1994).

Four well-known macro funds—Quantum Fund, Tiger Management, Steinhardt Partners, and Ardsley Partners—had net assets of almost $25 billion mid-1994 (“Fall Guys?” (1994)).

These possibilities are established formally by Kyle (1989).

These ideas are formalized by Pagano (1989a) and (1989b).

It is reported (e.g., Ito (1992)) that trust accounts of Japanese banks are much like mutual funds, and these data are not included in Tables II.6II.7. The size of trust accounts for Japanese banks in 1993 was $1.89 trillion, up from $1.56 trillion in 1992. The share of total assets in trust accounts that are foreign securities is similar to the numbers reported in Table II.7 for mutual funds—8.9 percent in 1993, down slightly from 9.3 percent in 1992.

In theory, these country weights should be based on all assets (stocks, bonds, real estate, and so on). A common simplification is, however, simply to use stock market capitalization.

Justification for this might be that the world market portfolio is a good rule of thumb for individual investors and that institutional portfolios should mirror underlying investors’ preferences. Some reasons are discussed below why institutional portfolios may be biased to domestic assets.

See, for example, Money Market Directories (1995). This prediction matches closely that of the International Finance Corporation (1994).

For instance, domestic ownership on the five largest stock exchanges is very high: 92 percent in the United States, 96 percent in Japan, 92 percent in the United Kingdom, 79 percent in Germany, and 89 percent in France (see French and Poterba (1991)).

For instance, see Davis (1991), Chuhan (1994), and Gooptu (1993).

The reason that these sorts of contracts are entered into could be explained as a response to other incentives and prudential problems. For instance, a defined benefit pension plan typically requires that the fund return some specific rate of return over some planning horizon.

Davis (1991) argues that it may be optimal for life insurance companies to concentrate their portfolios in domestic assets because matching currencies of assets and liabilities may be effective in limiting insolvency risk. This argument is essentially that shortfall risk may induce a home-asset preference.

It should be noted that the definition of the category “emerging markets” differs slightly across institutions. For instance, Singapore is not included in the IFC’s category, but it is included in the IMF’s balance of payments statistics. Emerging markets share of world GDP in 1993 was 20 percent. World market capitalization was $14.1 trillion in 1993; $12.5 trillion was accounted for by developed countries, of which $5.2 trillion was the U.S.; Mexico accounted for 1.4 percent of world market capitalization in 1993 (International Finance Corporation (1994)).

For instance, this view is held by Baring Securities. Howell (1993) holds that around 90 percent of flows to emerging markets are attributable to pension, insurance, and mutual funds.

The one-year surge in portfolio flows to Middle Eastern countries in 1988 is attributable to international bond issues by Bahrain ($80.6 million) and Israel ($20.0 million) (Organization for Economic Cooperation and Development, Financial Statistics Monthly, various issues).

International Finance Corporation (1994). For further analyses of the importance of U.S. and foreign mutual funds see the background paper “Capital Flows to Developing Countries,” pp. 33–52.

The Investment Company Act of 1940 stipulates that not more than 15 percent of the fund’s assets can be invested in illiquid assets—those not salable within seven days without a substantial discount. For further details, see “Fund Management” (1994).

Lipper Analytical Services.

The Investment Company Act of 1940 requires that redemptions be met on seven days’ notice.

Turnover ratios for open-end funds vary widely. Index funds, for instance, typically have low turnover ratios, whereas actively managed funds often have turnover ratios above 100 percent, and many “aggressive” funds have turnover ratios of several hundred percent. The average open-end fund has a turnover ratio of about 100 percent. Closed end funds typically have turnover ratios below 50 percent, and often in the neighborhood of 20 percent.

This estimate is higher than existing direct measures of emerging markets holdings of pension funds. For U.S. corporate pension funds, which account for just over 25 percent of the pension assets recorded in Table II.6, Money Market Directories (1995) puts emerging markets bond holdings in 1995 at 0.09 percent ($992 million) and emerging markets equities at 0.15 percent ($1,590 million). Interestingly, these measures are not much greater than Chuhan’s (1994) estimate for pensions funds’ holdings of emerging markets securities in 1992: she puts them at only 3.4 percent of total foreign asset holdings in 1992 (or $5.2 billion), which translates into an emerging markets position of less than 0.2 percent.

Richardson (1995).

Shiller (1981) used simple statistics and graphs to suggest that annual stock prices are more variable than is warranted by dividends. His methods were rigorously scrutinized by the profession, but his basic results have withstood many attacks. A survey of the work is given by Shiller (1991), see especially Chapter 4.

Short-run price movements mean a short period between successive price observations (fewer than 90 days, for example). The long-run refers to a longer sampling period (more than one year, for example).

See Fama and French (1988).

See Meese and Rogoff (1983).

The data are from International Financial Statistics for IFS 100-level countries.

Balassa (1964) discusses deviations from the theory caused by structural growth differences.

By contrast, the crash of October 29, 1929 caused a 24.5 percent reduction in the DJIA.

Readers can ignore technical aspects of this subsection without losing the logical flow of the arguments. The important fact to be kept in mind is that the basis of all analytical work involving bubbles is that the model in question lacks a unique equilibrium price.

This logic is developed by Tirole (1985).

In general the fundamental price is the currently expected present value of future dividends, and Pt=(ρ/(1ρ))·D¯ is a special case of that result, which is appropriate for discounting of constant dividends at the rate ρ.

In mathematical terms, this is because the line AA, whose slope is 1/ρ = 1/.95, is steeper than the 45-degree line.

This argument does not apply to models that are linear in the logarithm of price; a negative logarithm of price just means a price below unity.

The ideas worked out here for bubbles and fundamentals in a simple model of a stock market have counterparts in other models and markets.

Actually, it can only be concluded that the data might be inconsistent with the absence of bubbles.

The Mississippi Bubble, 1718–20, connected to John Law’s French government finance schemes and the temporally related South Sea Bubble, connected to the British government, are reviewed by Garber (1990).

Such work began with Flood and Garber (1980).

Two notable investigations were by West (1988) and by Froot and Obstfeld (1991).

See Ito (1992) for a general introduction to the asset price inflation in Japan. See also Schinasi and Hargraves (1993) for a description and an analysis of the asset price cycle in Japan and other industrial countries.

See Ito (1992) and (1993) and Ito and Iwaisako (forthcoming). Hoffmaister and Schinasi (1995) explain a significant part of land price movements in Japan in the late 1980s on the basis of macroeconomic fundamentals.

Much of this work involves adaptive expectations, which is a simplified error-learning mechanism that was popularized in Cagan (1956). In Cagan’s model, and those based on his work, if expectations adapt sufficiently quickly then price will be driven away from the long-run fundamental value instead of being driven toward it.

This market lasted a fixed number of “trading periods” (15–30), also known to the experimental subjects. The real time market exists for only one to two hours, so that discounting of future payouts is not important.

The “no-trade” theorems are developed by Grossman and Stiglitz (1980) and by Milgrom and Stokey (1982).

The presence of noise traders can, however, influence the fundamental equilibrium by injecting additional risk.

See, for example, Friedman (1953).

This becomes much more complicated when investor A tries to account for the effects of his actions on the decisions of investor B, who, in turn understands A’s motives.

    Other Resources Citing This Publication