The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.
Business Review Federal Reserve Bank of Philadelphia July • August 1996 ISSN 0007-701 1 Business Review The BUSINESS REVIEW is published by the Department of Research six times a year. It is edited by Sarah Burke. Artwork is designed and produced by Dianne Hallowell under the direction of Ronald B. Williams. The views expressed here are not necessarily those of this Reserve Bank or of the Federal Reserve System. SUBSCRIPTIONS. Single-copy subscriptions for individuals are available without charge. Insti tutional subscribers may order up to 5 copies. BACK ISSUES. Back issues are available free of charge, but quantities are limited: educators may order up to 50 copies by submitting requests on institutional letterhead; other orders are limited to 1 copy per request. Microform copies are available for purchase from University Microfilms, 300 N. Zeeb Road, Ann Arbor, MI 48106. REPRODUCTION. Perm ission must be obtained to reprint portions o f articles or whole articles. Permission to photocopy is unrestricted. Please send subscription orders, back orders, changes o f address, and requests to reprint to Publications, Federal Reserve Bank o f Philadelphia, Department o f Research and Statistics, Ten Independence Mall, Philadelphia, PA 19106-1574, or telephone (215) 574-6428. Please direct editorial communications to the same address, or telephone (215) 574-3805. 2 JULY/AUGUST 1996 REPEALING GLASS-STEAGALL: THE PAST POINTS THE WAY TO THE FUTURE Loretta J. Mester Passed as part of the National Bank Act of 1933, the Glass-Steagall Act prohibits the mixing of commercial and investment bank activities. It was passed during a time of tumult in financial markets: the economy was in depression and there were many bank failures. Given the state of today's banking industry and the cur rent economic climate, is it time to repeal Glass-Steagall? Congress has been debat ing the issue for some time. In this article, Loretta Mester weighs in with her analy sis of the situation. Her conclusion? The data support repeal. VALUE AT RISK: A NEW METHODOLOGY FOR MEASURING PORTFOLIO RISK Gregory P. Hopper Many different types of institutions hold portfolios of assets, and prudent financial management dictates that these firms be alert to any risks these assets may carry. How can these institutions judge the like lihood and magnitude of potential losses on their portfolios? A new methodology called value at risk (VAR) can be used to estimate these losses. In this article, Greg Hopper describes the various methods used to calculate VAR, paying special at tention to its weaknesses. FEDERAL RESERVE BANK OF PHILADELPHIA Repealing Glass-Steagall: The Past Points the Way to the Future Loretta ]. Mester* I n many countries, commercial banks are al lowed to perform investment banking activi ties such as helping their corporate customers bring new debt and equity issues to market. Yet in the United States, since the Glass-Steagall Act w as passed in 1933, m ost U.S. com m ercial banks are not permitted to engage in such un derwriting. Congress is debating whether to repeal this act. The legislation has undergone several revisions: some versions advocated al *Loretta M ester is vice president and econom ist and head of the Banking and Financial Markets section in the Research Departm ent of the Philadelphia Fed. lowing commercial banks to affiliate with in vestment banks in the same holding company, and som e advocated allow ing com m ercial banks to directly underwrite securities. While passage has seemed probable at many points, the measure has stalled. But this has more to do with provisions concerning com m ercial banks' right to sell insurance than with the pro posed repeal of the separation between com mercial and investment banking. Should Glass-Steagall be repealed? Bankers argue that the economic environment in which they operate has become much more competi tive and that they will fall behind unless they are permitted to expand their set of profitable activities, including investment banking. And 3 BUSINESS REVIEW it might be more efficient for commercial banks to engage in underwriting, since banks already have much information about their corporate customers. If so, society would gain from hav ing commercial banks engage in investment banking, since they could do it efficiently. On the other hand, as was argued at the time GlassSteagall was passed, there are potential conflicts of interest between commercial banking and underwriting. Whether these conflicts of in terest are present and whether they impose costs that outweigh the potential benefits of com m ingling investm ent and com m ercial banking activities is an empirical question. Several studies have sought evidence on this question by looking at the experience of banks before 1933, when they were allowed to under write securities with few restrictions; one study looks at more recent experience. The results suggest that conflicts of interest were not a major problem and still aren't—they support repeal of the Glass-Steagall Act of 1933. The studies present mixed results on whether it is better to have a commercial bank directly un derwrite securities or to house the commercial banking activities and investment banking ac tivities in separate subsidiaries of a holding company. THE ORIGINS OF GLASS-STEAGALL One of the main activities of an investment bank is underwriting. When a firm wishes to is sue new debt or equity, it goes to an investment bank, which prepares the issue. In underwrit ing, the investment bank usually guarantees to the firm that it will sell the issue at a specified price, which the bank determines after a credit evaluation of the firm and an assessment of market conditions. If the issue cannot be sold at the guaranteed price, the underwriter incurs the loss. This loss could occur because an un foreseen event causes the price of the issue to change during the period in which the under writer is trying to distribute the issue or because buyers have a different view of the firm's value 4 JULY/AUGUST 1996 than the underwriter did. Thus, to limit its own risk exposure, a good underwriter will need to know a lot about the firm and the firm's mar ket and be able to certify to the market that its assessment of the firm's value is correct. Prior to passage of the 1933 Glass-Steagall Act, state banks that were not members of the Federal Reserve System were permitted to un derwrite securities and bonds. The McFadden Act of 1927 allowed national banks to under write bonds, and they were later allowed to underwrite certain equity issues. But even be fore 1927, national banks engaged in securities activities by organizing state bank affiliates.1 So by the early 1920s, many commercial banks were heavily involved in the underwriting and distribution of securities.2 The number peaked in 1928 when 591 commercial banks were en gaged in securities activities either directly or through securities affiliates; of these, 235 were national banks and 356 were state-chartered.3 The background against which the GlassSteagall Act was passed was one of tumult in financial markets. The economy was in depres sion; there was a record number of bank fail ures. To the average person, it appeared the stock market crash had caused the Great De pression, and banks had had a large role in the stock markets. This perception, coupled with widespread bank failures, led Congress to be gin a series of investigations into market abuses and ways to reform the banking system, includ *Two companies are affiliates of one another if they have a com m on owner. One com pany is a subsidiary of another company if it is owned by that other company. 2For a review of the early history of bank securities ac tivities, see Benston (1990 and 1996) and Kroszner (1996). 3See Kroszner and Rajan (1994). Of course, in percent age term s, the number of banks engaged in securities ac tivities w as quite small— around 2.5 percent— since there were nearly 25,000 comm ercial banks in 1928 (see Board of Governors of the Federal Reserve System, N ovember 1943). FEDERAL RESERVE BANK OF PHILADELPHIA Repealing Glass-Steagall: The Past Points the Way to the Future Loretta /. Mester ing the famous Pecora hearings of the U.S. Sen banks have gradually added some investment ate in 1933-34.4 * banking activities to their portfolio of permis Congress was concerned about certain ques sible products. Today, commercial banks can tionable activities by banks and their securities perform agency functions for individual clients, affiliates. These activities included loans made that is, act as the client's agent in the market, by banks to their securities affiliates, loans ex including buying and selling stocks, safekeep tended by banks to others who wanted to buy ing securities, and switching funds between securities from the banks' securities affiliates, bank accounts and stock accounts. They can banks' buying securities underwritten by their operate discount brokerages, through which the affiliate for their own or their customers' ac public can buy stocks, and act as private place counts, and securities affiliates buying the stock ment agents (an issue marketed only to a few of firms that were custom ers of the bank. sophisticated investors is a private placement Rather than restrict these specific activities, and legally is not a security). They can advise Congress chose to separate commercial and clients on mergers. They can underwrite and investment banking altogether by passing the deal in municipal general obligation bonds, U.S. Glass-Steagall Act, which comprises four sec government bonds, Eurobonds (i.e., bonds is tions (16, 20, 21, and 32) of the National Bank sued outside the U.S.), m unicipal revenue bonds, and asset-backed securities. Some banks Act of 1933. Sections 16 and 21 prevent any bank that can underwrite and deal in corporate debt and accepts deposits from directly engaging in most equities, at least to a certain extent. They've been able to do this without violat securities activities except for those involving municipal general obligation bonds, U.S. gov ing Glass-Steagall by arguing that the different ernment bonds, private placements of commer language in Sections 20, 21, and 32 of the act— cial paper, and real estate bonds; these four are namely "engaged principally" in Section 20, called "eligible securities." Sections 20 and 32 "engaged" in Section 21, and "primarily en address indirect securities activities through gaged" in Section 32—indicates Congress es bank subsidiaries or affiliates and apply to tablished different standards for determining banks that are members of the Federal Reserve compliance with each of the provisions. Since System (which includes all national banks and the three terms weren't defined in the act, it has state-chartered banks that choose to become been up to the courts and regulators to deter members). Section 20 prohibits these banks mine the meaning and see that banks comply. from affiliating with any organization "engaged In a series of orders beginning in December principally" in underwriting securities, and 1986, the Federal Reserve stated that subsid Section 32 prohibits director, officer, or em iaries of bank holding companies set up to un ployee interlocks between these banks and derwrite U.S. government securities (which firms "primarily engaged" in securities activi were always "eligible" securities under GlassSteagall) may underwrite certain "bank ineli ties. gible" securities (the securities not included in the original four that Glass-Steagall allowed THE EROSION OF GLASS-STEAGALL Since Glass-Steagall was passed, commercial banks to underwrite) without violating Section 20 as long as the revenues obtained from un derwriting these ineligible securities were within certain limits. (See the Appendix: A Time 4The Stock Exchange Practices hearings of the Senate Line o f Permissible Securities Activities.) Committee on Banking and Currency were chaired by Sena tor Ferdinand Pecora. See Benston (1990) for discussion. 5 BUSINESS REVIEW JULY/AUGUST1996 ARGUMENTS FOR AND AGAINST REPEAL OF GLASS-STEAGALL Despite the fact that banks have been per mitted to engage to some extent in the under writing of corporate debt and equity and other ineligible securities, these activities are still highly regulated. Banks wishing to underwrite ineligible securities must seek approval from the Fed to set up so-called "Section 20" affili ates, and the revenue limits placed on such underwriting have begun to become binding on some banks.5 As of March 31, 1996, in the U.S. there were 38 Section 20 subsidiaries of commercial bank holding companies autho rized to engage in limited underwriting of and dealing in ineligible securities, including mu nicipal revenue bonds, 1-4 family conventional mortgage-backed securities, commercial paper, and asset-backed securities (Table l).6 Twentyone of these subsidiaries were authorized to un derwrite both corporate debt and equity secu rities, and an additional three were authorized to underwrite corporate debt but not equities. Most of these organizations are located in the New York Federal Reserve District, where they can directly compete with large investment banks. Arguments for Repeal. The erosion of GlassSteagall suggests that commercial banks have had strong incentives to get into securities ac tivities and that regulators have had incentives to allow the banks to do so. One reason banks want to perform these activities is that they are profitable. As financial markets have become deregulated, banks have faced increased com petition for their core businesses of deposit-tak ing and loan-making. In addition, technologi cal advances have allowed firms increased ac cess to funding from nonbank sources. Thus, finding new pathways to profits has become increasingly important for commercial banks, and it appears that underwriting securities is one such avenue. For example, in 1993, the average return on equity for large investment banks was over 23 percent; for New York Stock Exchange broker/dealer firms it was about 16.25 percent; and for commercial banks it was about 15.25 percent.7 For the period 1990 to 1993, the return on equity averaged about 17.5 percent for investment banks and about 11 per cent for commercial banks. In addition, commercial banks that could also offer unlimited underwriting services would be able to retain some of their most cred itworthy customers. These customers usually find it cheaper to issue commercial paper than to take out bank loans, so they have been turn ing to the markets to raise funds. Because of the legal limits on the amount of commercial paper commercial banks are permitted to un derwrite, they have lost some of their better customers. This loss could lead to a contrac tion in the banking industry, which could im pose costs on smaller, less creditworthy firms that cannot access the markets directly but de pend on bank loans for financing. One can also make the argument that allow ing commercial and investment banking activi ties within the same institution could make the industry safer by allowing more diversifica tion.8 In addition, there could be natural syn ergies between commercial and investment 5H ays and Wilke (1996) and Rehm (1994) discuss banks hitting the revenue limit. The Federal Reserve recently fined Swiss Bank Corporation $3.5 million for exceeding the rev enue limit. 7Data from the Board of Governors of the Federal Re serve System, and from the FDIC Quarterly Banking Profile, FDIC, Third Q uarter 1995. 6Some are foreign owned; one of the subs has been dor mant since June 1995; one holding com pany has two Sec tion 20 subs. 6 8But there is m ixed evidence on w hether mixing com mercial banking with other nonbank activities leads to lower insolvency risk for the institution. See M ester (1992a) for a brief literature review of the empirical evidence. Loretta ]. Mester Repealing Glass-Steagall: The Past Points the Way to the Future TABLE 1 Section 20 Subsidiaries Banking organizations authorized to underw rite and deal in certain m unicipal revenue bonds, mortgagerelated securities, com m ercial paper, and asset-backed securities as of M arch 31, 1996, listed by Federal Reserve District. Date of Initial Board Order Authorization Date of Initial Board Order Authorization Boston District Fleet Financial Group 10/88 N ational City Corp.3 PNC Bank Corp. New York District Banco Santander, S.A .a The Bank of Nova Scotia3 Bankers Trust N.Y. Corp.3 Barclays Bank PLCb Canadian Im perial Bank of Com m erce3 Chase M anhattan Corp.3 Citicorp3 Deutsche Bank A G a Cleveland District (continued) H S B C H o ld in g s P L C a 3/95 4/90 4/87 1/90 1/90 5/87 4/87 12/92 2/96 The Long-Term Credit Bank of Japan, Ltd. J.P. M organ & Co.3 The Royal Bank of Canada3 Saban/Republic New York Corp.3 Swiss Bank Corp.3 The Toronto-Dom inion Bank3 5/90 4/87 1/90 1/94 12/94 5/90 ABN AM RO Bank N.V.a d The Bank of M ontreal3 First of Am erica Bank Corp.b First Chicago N BD Corp.b 6/91 Norw est Corp. 7/90 12/92 2/96 4/95 BankAm erica Corp.3 D ai-Ichi Kangyo Bank Ltd. The Sanw a Bank, Ltd. 8/89 5/89 Atlanta District Barnett Banks, Inc.c SouthTrust Corp. SunTrust Banks, Inc. Synovus Financial Corp. 1/89 7/89 8/94 9/91 Chicago District 6/90 5/88 10/94 8/88 12/89 San Francisco District Cleveland District Bank One Corp. H untington Bancshares, Inc. KeyCorp M ellon Bank Corp. Richmond District First Union Corp.3 N ationsBank Corp.3 Minneapolis District Philadelphia District D auphin Deposit Corp.3 2/94 7/87 3/92 1/91 5/90 aAlso has corporate debt and equity securities powers. bAlso has corporate debt securities powers. cAs of June 3 0 ,1 9 9 5 , the Section 20 subsidiary w as dormant. dHas two Section 20 subsidiaries. Source: Various issues of the Federal Reserve Bulletin. 7 BUSINESS REVIEW banking. For example, credit evaluation is im portant in both. Loan syndication, which is permitted for commercial banks, is very simi lar to underwriting. And banks are already experienced at underwriting eligible securities. There may be scope economies from reusing information from the credit evaluation of a bor rower who subsequently wants to issue debt.9 Commercial banks obtain valuable (inside) in formation on their customers from monitoring their loans: they see the firms' payment history and cash flows. So when the issuing firm is a customer of a commercial bank, the informa tion this bank would have if it were to under write the issue is likely to be more accurate than the information an investment bank under writer would have. Thus, it might be more ef ficient having commercial banks engage in un derwriting than having specialized investment banks do it— if so, society would gain. Arguments Against Repeal. These poten tial benefits have to be weighed against the po tential costs stemming from possible conflicts of interest between commercial banking and underwriting.1 (If there are no costs, one could 0 argue for repeal even if potential benefits are meager.) Some of these conflicts were raised during the Pecora hearings. A commercial bank might promote the securities it underwrites and misrepresent the quality of these securities to its depositors instead of offering them disinter ested investment advice. Or the bank might induce a troubled loan customer to issue new securities to repay the loan. This imposes costs. If investors in these securities are naive, they are penalized: they purchase poor quality se curities thinking they are good. If, however, investors are not naive, they know such a con 9But, again, there is some empirical evidence that sug gests this m ay not be the case. See Mester (1992b). 10Saunders (1985), Saunders and W alter (1994), and Walter (1985) also discuss conflict-of-interest arguments against repeal of Glass-Steagall. 8 JULY/AUGUST 1996 flict of interest might exist and will, therefore, adjust down the price they are willing to pay for such securities. In this case, the issuing firms that use commercial bank underwriters bear the cost: they receive less funding than they would like, so there is u n d erin v estm en t.1 The 1 economy is worse off, since some good invest ments go unfunded. A cost is also imposed on c o m m e rc ia l b a n k s th a t w a n t to d e v e lo p re p u tations for good underwritings. These potential costs from a possible conflict of interest have to be weighed against the po tential benefits of allowing commercial banks to underwrite. There is a trade-off: a commer cial bank may obtain needed information more efficiently than an investment bank, but it may misrepresent this information to the market. An investment bank doesn't have ties to the issuer, so it has less incentive to misrepresent the in formation, but its information may not be as accurate. Whether the information cost savings of a commercial bank underwriter outweigh the costs imposed by the potential conflict of inter est is an empirical question. EMPIRICAL EVIDENCE ON CONFLICTS OF INTEREST If conflicts of interest presented problems, such problems should have manifested them- 1 While the underw riter bears the cost if it guarantees a 1 high price to the issuer and can obtain only a low price when selling the securities to investors, a sm art comm ercial bank underwriter would take into account the sm art investors' downward price adjustment and not guarantee a high price to the issuer. Hence, it's the issuer that bears this cost, and society, since some good investment projects go unfunded. To the extent that a firm could switch to an investment bank underwriter, this underinvestment problem w ould go away. But switching m ight be difficult because the market might not be able to determine w hether a firm w as switch ing to avoid the underpricing problem or because its com mercial bank refused to underw rite the firm's securities because they were not of high enough quality (see Rajan, 1996). The underinvestment problem could also be avoided if firms used only investment banks to underw rite their se curities. FEDERAL RESERVE BANK OF PHILADELPHIA Repealing Class-Steagall: The Past Points the Way to the Future selves in the period before Glass-Steagall was enacted. Yet empirical studies that examine the 1920s and early 1930s suggest that conflicts were not generally a problem, and a study of the modern securities activities of commercial banks suggests they still aren't.1 (Note, how 2 ever, that finding no conflict of interest is not the same thing as finding benefits to allowing commercial banks into underwriting.) Actual Performance. If banks systematically underwrote poorer quality security issues and passed them off to their depositors, the issues underw ritten by com m ercial banks would probably have performed worse than similar issues underwritten by investment banks over the same period— that is, the measures of ac tual performance would differ according to the underwriter. Also, if the public had been taken advantage of in this way, it probably would have been easier to do with issues of low-qual ity and lesser-known firms, about which little public information was circulating. But three interesting studies all found evidence that se curities underwritten by commercial banks ac tually outperformed those underwritten by in vestment banks in the pre-Glass-Steagall pe riod. Manju Puri (1994) studied the default per formance and mortality rates (default rates ad justed for the ages of the issues) of a sample of securities issued over the period January 1927 to September 1929, when national as well as state banks were authorized to underwrite bonds (Table 2). In comparing the default per formance of the issues she not only distin 12Bank of United States is often cited as an example of a bank that failed because of its affiliates' abusive practices. But as Benston (1996) notes, only one of these affiliates dealt in securities, and it w as engaged in purchasing the bank's stock, not in underw riting other firms' securities. The rest were involved in real estate. Benston cites rapid expansion and m isappropriation of funds by the bank's owners as the chief reasons for the bank's failure. Loretta J. Mester guishes between issues underwritten by com mercial banks and investment banks (which she calls nonbanks), but also issues underwritten by National City Company and Chase Securi ties Corporation. These so-called rogue banks were accused of abuses and investigated by Congress in hearings surrounding GlassSteagall. She also considers the type of secu rity underwritten and whether the issue was investment or noninvestment grade. Puri generally finds that the mortality rates for issues underwritten by commercial banks are significantly lower (in a statistical sense) than those underwritten by investment banks. For example, she finds that seven years after issue, about 25 percent of the industrial bonds underwritten by commercial banks had de faulted, while almost 40 percent of those un derwritten by investment banks had defaulted. She finds statistically significant differences for these bonds three years and five years after is sue as well, and significant differences even when the bonds were divided into investment and noninvestment grade issues. For preferred stock, the results are a bit weaker, perhaps be cause the sample size is smaller. Puri did not find a significant difference in mortality for for eign bond issues taken as a group, but she did find one for the noninvestment grade sub group. Perhaps not surprisingly, she finds that issues underwritten by the rogue banks gener ally defaulted more than issues underwritten by the other banks. She didn't report a statisti cal test, but her estimates suggest that, at least for the older issues, rogue bank issues defaulted more than investment bank issues. James Ang and Terry Richardson (1994) stud ied a sample of 669 domestic and foreign cor porate and foreign government bonds under written from 1926-34 and obtained results simi lar to those of Puri. They studied the default experience of these issues from the time of is sue until 1939 and found that commercial bank underw ritings significantly outperformed those of investment banks: about 40 percent 9 TABLE 2 Empirical Studies Study Time Period Sample Selected Results Puri (1994) January 1927 to September 1929 Samples ranged in size from 365 to 382 issues. Default experience over the seven years after issue was available for 181 industrial bonds, 81 preferred stock issues, and 103 government bonds. Default experience over the year after issue was available for 182 industrial bonds, 95 preferred stock issues, and 105 foreign government bonds. In the larger sample, 134 issues were underwritten by commercial banks and 248 were underwritten by investment banks. Issues underwritten by commercial banks defaulted less often than issues underwritten by investment banks, Puri (1996) Same as above. Same as above. Issues underwritten by commercial banks had lower initial yields than issues underwritten by investment banks. Compared to issues underwritten by investment banks, issues underwritten by commercial bank affiliates had similar initial yields while issues underwritten directly by commercial banks had lower initial yields. Ang and Richardson (1994) 1926-34 669 domestic and foreign corporate bonds and foreign government bonds. 121 were underwritten by commercial banks, 451 were underwritten by investment banks, and 97 were underwritten by Kuhn, Loeb and Co. or J.P. Morgan. Issues underwritten by commercial banks defaulted less often and had lower initial yields than issues under written by investment banks. Kroszner and Rajan (1994) First quarters 1921-29 462 industrial bonds. 133 were underwritten by commercial banks and 329 were underwritten by investment banks. Used to form 121 matched pairs. Issues underwritten by commercial banks defaulted less often and had lower initial yields than issues underwrit ten by investment banks. Commercial banks were more likely to underwrite issues of larger, older, and less leveraged firms, firms listed on the stock exchange, and senior securities. Kroszner and Rajan (1995) 1925-29 906 issues of common and preferred stock and corporate and government bonds underwritten by commercial banks. 580 were underwritten by commercial bank affiliates and 326 were underwritten directly by commercial banks. Initial yields on issues underwritten by commercial bank affiliates were lower than initial yields on issues underwritten directly by commercial banks. Gande, Puri, Saunders, and Walter (1995) January 1, 1993 to March 31, 1995 670 fixed-rate, nonconvertible debt issues of nonfinancial corporations. 80 were underwritten by Section 20 affiliates of commercial banks and 590 were underwritten by investment banks. Initial yields on issues underwritten by Section 20 subsidiaries of commercial banks and by investment banks were generally the same, but initial yields on issues with low credit ratings whose proceeds were not being used to repay issuer's bank loans were lower when underwritten by Section 20 subsidiaries. Repealing Glass-Steagall: The Past Points the Way to the Future Loretta ]. Mester defaults compared with more than 48 percent defaults for the investment bank issues; com mercial bank issues outperformed investment bank issues for each type of security examined. They also found that the issues underwritten by Kuhn, Loeb and Co. and J.P. Morgan, insti tutions that were difficult to classify as either commercial or investment banks, outperformed both commercial and investment bank issues, with a default rate of only 30 percent. But even including these two institutions among invest ment banks does not change the result that com mercial bank underwritings defaulted less of ten than investment bank underwritings. In their study, National City and Chase did worse than other commercial banks, but they seem to have been on a par with investment banks. Randall Kroszner and Raghuram Rajan (1994) conducted a matched-sample test of 121 pairs of industrial bonds underwritten during the first quarters of 1921-29. The bonds in each pair were matched in terms of their initial rat ing, time when issued, maturity, size, and type of conversion provisions, but one bond in the pair was underwritten by a commercial bank while the other was underwritten by an invest ment bank.1 Again, their results agree with 3 those of the other studies: they find that at the end of every year after 1924, fewer cumulative defaults occurred among the issues underwrit ten by commercial banks than among those underwritten by investment banks. By 1940, 32 percent of investment-bank underwritings had defaulted compared with 23 percent of commercial bank underwritings. Thus, these three studies of the performance of issues found no evidence that commercial banks were foist ing off low-quality securities on investors. Expected Performance. While the actual performance of these issues is important, so is the expected performance. Only if, on average, default of the issues was greater than expected can one conclude that investors were being duped by the underwriter. Evidence on this can be garnered by looking at the pricing of the issues. Studies by Ang and Richardson (1994) and Puri (1996) found that securities underwrit ten by commercial banks were priced higher (that is, their yields were lower) at the time of issue than securities underwritten by invest ment banks, meaning that investors did not require that a high risk premium be built into the yield to induce them to buy commercial bank issues. For example, Ang and Richardson found that over 1926-30 the initial yield on issues under written by commercial banks averaged about 26 basis points lower than the yield on issues underwritten by investment banks.1 Appar 4 ently investors did not perceive that issues un derwritten by commercial banks were neces sarily more risky than those underwritten by investment banks. The study also found that the actual yield performance (that is, the return over the life of the issue) of the issues under written by commercial banks was better than that of investment bank issues, which is con sistent with the default rate results discussed above. Moreover, Ang and Richardson also per formed a statistical test to shed some light on whether investors were rationally assessing the value of the issues. If they were, the yield at the time of issue should be a good predictor of the realized yield of the issue. Ang and 13Their definition of a comm ercial-bank-underwritten issue w as broader than Puri's. A n issue w as classified as a comm ercial bank underwriting if a commercial bank w as a member of the group of institutions, that is, the syndicate— either as a lead or subordinate mem ber— that underwrote the issue. Puri classified an issue as a commercial bank underwriting only if a comm ercial bank w as the sole un derw riter or the lead underwriter, arguing that subordinate m em bers of a syndicate could exert only a limited amount of influence on other members. 14A basis point is 1 / 100th of a percent. 11 BUSINESS REVIEW Richardson found no evidence that the market mispriced issues underwritten by commercial banks and no evidence that the predictive power of the issue price for realized yield was different for commercial-bank-underwritten issues than for investment-bank-underwritten issues. Hence, they found no evidence that in vestors were behaving irrationally when they accepted lower yields for the commercial bank issues. Examining the same sample of issues as in her previous study, Puri (1996) also found that investors were willing to pay higher prices (that is, accept lower yields) for securities underwrit ten by com m ercial banks than investm ent banks, after controlling for other factors that would have affected prices.1 This result held 5 for both industrial bonds, where the yield at the tim e of issue on com m ercial bank underwritings averaged between 8 and 13 ba sis points lower than that on similar investment bank underwritings (depending on the statis tical methodology used), and for preferred stock issues, where the difference was between 22 and 37 basis points.1 6 One interpretation of this result is that in vestors assessed that the commercial bank's potential information advantage over the in vestment bank outweighed any potential con 15These factors included w hether the issue w as invest ment grade, the size of the issue, the size of the underw rit ing syndicate, whether the issue was traded on an exchange, the firm's age, and w hether the firm had issues of the same type (either bonds or preferred stock) outstanding in the market. Puri used this last factor to define w hether the is sue was a new or seasoned issue, since it w as not possible from the available data to determine w hether an issue w as the firm's first ever issue of a bond or preferred stock. 16Given that comm ercial bank underw riters appear to have been able to generate higher prices for their issuers, there is a question as to w hy any issuer would have chosen an investment bank as underwriter. Puri (1996) suggests that one reason m ight be that investment banks charged lower fees (although she has no data on fees to confirm this conjecture). 12 JULY/AUGUST1996 flict-of-interest problem in the commercial bank. Hence, they were willing to pay higher prices for issues underwritten by commercial banks. If so, issuers did not bear the costs of potential conflicts of interest. Consistent with this interpretation is Puri's finding that the dif ference in yields for commercial and investment bank issues was greater for new issues (that is, issues different from the type, either bonds or preferred stock, that the firm had outstanding in the market) than for seasoned issues (that is, issues that were similar in type to ones the firm had outstanding in the market). Typically there is less public information available on new is sues, so any private information a commercial bank has should be more valuable for new is sues than for seasoned issues. So if the market believes the commercial bank has an informa tion advantage over the investment bank in underwriting, and this influences the prices it is willing to pay for securities, one would ex pect to see a larger price difference between commercial bank underwritings and invest ment bank underwritings for new issues than for seasoned issues, which is what Puri found. She also found no yield differential in foreign bond underwritings. Since prior lending rela tionships were not important in gaining cus tomers in this market, there was little reason to believe a commercial bank's information was superior to that of an investment bank under writer. Types of Issues. The default and price re sults are based on a comparison of issues un derwritten by com m ercial and investm ent banks that are similar in other respects so that any differences found can be attributed to un derwriter type. For example, the studies com pare securities of similar types, with the same maturities, size, and so on. But the studies also found that the general types of securities un derwritten by com m ercial and investm ent banks differed. Puri (1996) found that commer cial banks were more likely to underwrite cor porate bonds than preferred stock, and of the FEDERAL RESERVE BANK OF PHILADELPHIA Repealing Glass-Steagall: The Past Points the Way to the Future corporate bond issues, they were more likely to underwrite seasoned issues, those of older firms, those with less underlying collateral se curing the issue, and those with a larger num ber of underwriters in the syndicate. Kroszner and Rajan (1994) found that commercial banks were more likely to underwrite larger and older firms, firms listed on the stock exchange, less leveraged firms, and senior securities such as debt rather than stock. These characteristics are generally consistent with higher quality issues. Moreover, they found that these differences were more pronounced for smaller banks than for larger ones. Kroszner and Rajan argue that one explana tion for their findings is that commercial banks were deliberately choosing to underwrite highquality issues, which involve less insider infor mation, and so have lower potential conflicts of interest. That is, commercial banks wanted to indicate to the market that they were cred ible underwriters, so they focused on the types of issues that minimized the risk of conflicts of interest. Since small banks, as relative un knowns, likely need to do more to build their reputations, Kroszner and Rajan's result show ing that small banks focused even more on high-quality issues than large banks did is con sistent with this explanation. However, other plausible explanations have little to do with conflicts of interest. For ex ample, it could be that commercial banks fo cused on debt securities rather than equities because they had more expertise with these types of securities. Recall that this was the type of security they were first authorized to under write, and debt securities are more like com mercial bank loans.1 7 Recent Experience. I know of only one study of the underwriting experience of com mercial banks since the Federal Reserve permit ted limited underwriting of ineligible securities. 17See Puri (1996) for further discussion. Loretta J. Mester It is still too early to determine the default ex perience of recent issues, but Amar Gande, Manju Puri, Anthony Saunders, and Ingo Walter (1995) were able to examine the pricing of issues underwritten by the top 20 underwrit ers (in terms of the dollar volume of their underwritings) of fixed-rate, nonconvertible debt issues of nonfinancial corporations over the period January 1,1993, to March 31,1995. This sample included four underwriters that are Section 20 subsidiaries of commercial bank holding companies: J.P. Morgan, Citicorp, Bank ers Trust, and Chase. In addition to isolating a commercial bank's corporate debt and equity underwriting activi ties in a separate affiliate of the commercial bank within the holding company, regulators impose firewalls that limit the financial and information flows between the securities and commercial bank subsidiaries. Firewalls are in tended to stop the conflict-of-interest problem, but at the same time, they restrict the ability of commercial banks to take advantage of any in formational edge they may have in underwrit ing as a result of their lending activity. Gande, Puri, Saunders, and Walter studied the effectiveness of these firewalls by compar ing the pricing of similar issues underwritten by Section 20 subsidiaries and investment banks, while controlling for the lending rela tionship between the commercial bank under writer and the issuer. That is, the study goes a step further at getting at the conflict-of-interest problem by controlling for the volume of loans an issuer has gotten from the commercial bank affiliate of its Section 20 underwriter. (Recall that the potential conflict-of-interest problem should be worse when a commercial bank un derwriter has also extended loans to the issuer.) If firewalls have successfully prevented con flicts of interest and have precluded the com mercial bank from taking advantage of any in formational edge it might have over the invest ment bank underwriter, one would expect to see no difference in the initial yields of similar 13 BUSINESS REVIEW JULY/AUGUST1996 Section 20 and investment bank underwritings. One would also expect to see no yield differ ence if there weren't any conflicts of interest or informational advantages in the first place, or if the conflicts of interest just offset the infor mational advantages, regardless of the effec tiveness of firewalls. On the other hand, if the market assesses that the informational advan tages of the commercial bank underwriter out weigh any conflicts of interest and that the firewalls are not fully effective at isolating the underwriting function from the commercial banking function, yields on issues underwrit ten by Section 20 subsidiaries should be lower than those on investment bank underwritings. Similarly, if the market assesses that the po tential conflict-of-interest problem outweighs any potential informational advantage com mercial banks have in underwriting and that the firewalls are not fully successful in control ling conflicts of interest, the market should re quire a higher risk premium to take on com mercial bank underwritings. Thus, initial yields on Section 20 subsidiary underwritings should be greater than initial yields on investment bank underwritings. The authors found no statistically significant difference, on average, in the yields of issues underwritten by Section 20 subsidiaries and similar issues underwritten by investment banks.1 Thus, it appears that, on average, ei 8 ther firewalls have been effective at isolating the underwriting and commercial bank func tions or that any informational advantages just offset any conflicts of interest. However, they also found that when a Section 20 subsidiary underwrites issues whose proceeds are not in tended to repay the issuer's bank loans and the issue has a low credit rating, the yield at the time of issue is significantly lower than if an investment bank underwrites the issue. These issues are likely to present the fewest conflictof-interest problems, since the proceeds of the issue are not being used to repay a loan and thereby shift the risk out of the bank underwriter's loan portfolio on to those who purchase the underwritten issues. And any informational advantage that a commercial bank underwriter has is likely to be most valu able with lower rated issues. The fact that the market accepts a lower yield suggests that for this type o f security issue the market believes that the commercial bank's information advantage outweighs the cost from any conflict of interest and that firewalls are not fully isolating the underwriting function from the commercial banking function.1 It is too early to tell whether 9 these beliefs are rational, since it depends on the actual default experience of the issues. 18While they found that, on average, yields on Section 20 underwritings were low er than yields on investment bank underwritings, the difference was not statistically sig nificant. 19The authors also find no evidence of the conflict-ofinterest problem in those issues where one might expect the problem to be severe, namely, in issues whose proceeds are being used to repay bank loans. Digitized 14 FRASER for EMPIRICAL EVIDENCE ON ORGANIZATIONAL STRUCTURE While the empirical studies have been con sistent in suggesting that conflicts of interest have not been a major problem that should pre clude commercial banks from participating in securities activities, they provide mixed evi dence on the way these activities should be or ganized. The version of the repeal legislation that Congress has been considering in its cur rent session would allow commercial banks into securities activities through a holding com pany structure, the same structure that has been used since the Federal Reserve permitted lim ited securities activities in 1987. That is, rather than the commercial bank's engaging in the securities activities directly, the securities and commercial banking activities would be in separate subsidiaries of a financial service hold ing company. The separate affiliates would be FEDERAL RESERVE BANK OF PHILADELPHIA Repealing Glass-Steagall: The Past Points the Way to the Future Loretta J. Mester further protected by a system of firewalls. to September 1929, she found that the yields at This organizational structure with firewalls the time of issue on corporate debt issues un provides a benefit in lowering the potential for derwritten by affiliates did not differ signifi conflicts of interest and so lowers the costs com cantly from the yields on similar issues under mercial banks and issuers must incur to assure written by investment banks, but that the yields investors their issues are high quality. But it of direct underwritings were significantly less also imposes a cost by making the information (from 9 to 23 basis points lower, depending on sharing between the lending and underwriting the method of estimation) than those on invest functions more difficult. The study by Gande, ment bank underwritings. Similar results hold Puri, Saunders, and Walter indicates that, re for preferred stock issues. While a test that di cently, firewalls haven't always been effective rectly compares the yields on issues underwrit in totally separating the commercial bank and ten in-house with those on issues underwrit securities affiliates, but it also indicates that ten by an affiliate would be more definitive, conflicts of interest haven't been a problem. Puri's results do suggest that yields would be Two other studies that examined issues di low er on in -h ou se than on affiliate rectly underwritten by commercial banks and underwritings, given their respective relation those underwritten by an affiliate of a commer ships to yield s on in vestm en t bank cial bank in the pre-Glass-Steagall period came underw ritings. This is consistent with the up with conflicting conclusions as to which m arket's not b eliev in g that direct organizational structure is preferable. Kroszner underwritings were subject to greater conflicts and Rajan (1995) found that firewalls appear of interest than affiliate underwritings; other to have been valuable in helping commercial wise, purchasers would have demanded higher bank underwriters convince the market they yields on direct underwritings, not lower ones. were not trying to foist off poor-quality issues. Since Puri's conclusions differ from those of They studied 906 issues underwritten by com Kroszner and Rajan, perhaps because different m ercial banks between 1925 and 1929 and samples of security issues were studied, a de found that yields on issues underwritten di finitive answer on the issue of organizational rectly by banks averaged 13 basis points higher form awaits further study. than yields on similar issues underwritten by affiliates. This indicates that the market as CONCLUSIONS sessed that potential conflicts of interest were Congress has been debating whether to re higher with direct underwriting. They also peal the Glass-Steagall Act, which was passed found that, over the 1920s, banks increasingly in 1933 in the aftermath of the large number of organized their securities activities in affiliates bank failures that occurred during the Great rather then keeping them in-house. It seems Depression. One of the problems the act sought sensible that the market would have evolved to address was the potential conflict of interest this way, since the affiliates appeared able to when a commercial bank that lends to a firm guarantee higher prices to their issuing custom also underwrites that firm's securities. Empirical evidence based on the pre-Glassers. But P uri (1996) concluded that direct Steagall days and on commercial banks' recent underwritings by commercial banks did not experience in debt underwriting suggests that, lead to greater conflicts of interest than under on balance, conflicts of interest have not been a w riting via affiliates. With her sample of problem: the data support the repeal of Glassunderwritings over the period of January 1927 Steagall. 15 June 1988 APPENDIX: A Time Line of Permissible Securities Activities The Fed allowed subsidiaries of com m ercial banks to under write commercial paper, municipal revenue bonds, mortgagebacked securities (as long as they w eren't m ortgages of an affiliated bank), and securities backed by unaffiliated banks' consumer-related receivables, subject to the revenue restric tion. 1986 1987 1988 198 December 1986 The Federal Reserve issued a policy stating that governm ent securities subsidiaries of bank holding com panies may underwrite certain "bank ineligible" securities without violating Section 20 of the Glass-Steagall Act as long as the underw riting revenues from ineligible securities did not exceed 5 percent of the subsid iaries' gross revenues, since this would indicate that the subsidiary was not "engaged principally" in un derwriting ineligible securities. The revenue test had to be m et on an eight-quarter m oving average basis. Note that even though banks were always permitted to directly underw rite U.S. governm ent securities, if a bank w anted to underw rite ineligible securities in an affiliate, it made sense to m ove the underw riting of governm ent securities to the affiliate as well, since this would increase the gross revenues of the affiliate and, therefore, the volum e of ineligible securities the bank's affiliate could underwrite. To lim it the possi bility of conflicts of interest, the Fed included several "firew alls" (see M ester (1992a) for a discussion of these firewalls and their limitations): 1. Securities activities had to be in a subsidiary of the holding com pany that w as separate from the com mercial bank. These subsidiaries are called Section 20 subsidiaries. 2. Transactions betw een the affiliated bank and securities subsidiary were limited. 3. The securities and com m ercial bank subsidiaries could have no officers, directors, or em ployees in common. 4. The com m ercial bank subsidiary was restricted in extending loans to issuers of com m ercial paper placed by the securities affiliate. 5. The com m ercial bank subsidiary could not purchase or recomm end that its custom ers purchase securities placed by its securities affiliate. 6. The securities subsidiary could have only limited access to custom er records of the com m ercial bank subsidiary and could not underw rite securities issued by affiliates. September 13,1989 The Fed raised the limit on revenues from underw riting "bank ineligible" securities by Section 20 subsidiaries to 10 percent from 5 percent and allowed subsidiaries to under write and deal in securities issued by their affiliates if the securities are rated by a nationally know n rating agency or guaranteed by a governm ent agency like Fannie M ae, Freddie M ac, or Ginnie Mae. 1990 1991 September 20,1990 The Fed authorized the first bank holding com pany, J.P. M organ, to underw rite equities in a subsidiary of the holding company, subject to the 10 percent revenue limit. 1992 1993 1994 T J January 8,1989 July 1994 The Fed said Section 20 does not bar bank holding com pany subsidiaries from underw riting and deal ing in corporate debt and, after a waiting period (as short as a year later) in corporate equity. To be au thorized, the holding com pany must, among other things, be well capitalized. The Fed authorized five large bank holding com panies to underw rite corporate debt with the ruling: J.P. M organ, Chase, Bankers Trust, Citicorp, and Se curity Pacific. J.P. M organ Securities, the Section 20 subsidiary of J.P. M organ, Inc., did the first publicly issued cor porate bond underw riting by a com m ercial bank af filiate in January 1989. The Fed sought com m ent on a pro posal for an alternative to the revenue test that would lim it underw riting of ineligible securities to 10 percent of asset value of the subsidiary or of sales volum e, or both. The proposal is still alive but is on hold pending the final outcom e of Glass-Steagall reform legislation. January 1993 Several banks were reaching the 10 percent revenue lim it placed on securities activities. The Fed permitted an optional method for m eeting the limit: indexing the revenue test to interest rate changes. To account for changes in the level and slope of the yield curve since September 1989, the banks were allowed to calculate the revenue that would have been earned if the yield curve had been as it was in Septem ber 1989. The rationale for the change was that unusual changes in interest rates, w hich had occurred since 1989, had made the 10 percent revenue test more binding than it was w hen originally adopted. Namely, the spread between long and short rates had widened substantially. Since ineligible securities tend to be longer term than eligible securities, this m eant that the 10 percent revenue limit had become more stringent even for banks that had not changed the proportion of ineligible to eligible securities they underw rote (see the Federal Register, 1994). BUSINESS REVIEW JULY/AUGUST 1996 Selected Bibliography Ang, Jam es S., and Terry Richardson. “The Underw riting Experience of Comm ercial Bank Affiliates Prior to the Glass-Steagall Act: A Re-exam ination of Evidence for Passage of the A ct," Journal o f Banking and Finance 18 (January 1994), pp. 351-95. Benston, George J. “The O rigins and Justification for the Glass-Steagall A ct," in Universal Banking: Financial System Design Reconsidered. Hom ew ood, IL: Irw in (1996), pp. 31-69. Benston, George J. The Separation o f Commercial and Investment Banking. New York: O xford University Press, 1990. Board of Governors of the Federal Reserve System. Banking and Monetary Statistics 1914-1941. W ashington, D.C., N ovem ber 1943. Federal Register 59 (Thursday, July 12,1994), pp. 35516-19. Gande, Amar, Manju Puri, Anthony Saunders, and Ingo Walter. “Bank Underw riting of D ebt Securities: M odern Evidence," New York University m anuscript (November 1995). Hays, Laurie, and John R. Wilke, "Banks Bump Against Cap on D ealing," Wall Street Journal (M arch 29, 1996), p. C l. Kroszner, Randall S. "T h e Evolution of U niversal Banking and Its Regulation(s) in Tw entieth Century A m erica," in Universal Banking: Financial System Design Reconsidered Hom ew ood, IL: Irw in (1996), pp. 70-99. Kroszner, Randall S., and Raghuram G. Rajan. "Is the Glass-Steagall Act Justified? A Study of the US Experi ence w ith Universal Banking Before 1933," American Economic Review 84 (Septem ber 1994), pp. 810-32. Kroszner, Randall S., and Raghuram G. Rajan. "O rganization Structure and Credibility: Evidence from C om mercial Bank Securities Activities Before the Glass-Steagall A ct," N ational Bureau of Econom ic Re search Working Paper 5256 (Septem ber 1995). Mester, Loretta J. "Banking and Com m erce: A Dangerous Liaison?" Business Review, FederalReserve Bank of Philadelphia (M ay/June 1992a), pp. 17-29. Mester, Loretta J. "Traditional and N ontraditional Banking: An Inform ation-Theoretic A pproach," Journal o f Banking and Finance, 16 (1992b), pp. 545-66. Puri, Manju. "C om m ercial Banks in Investm ent Banking: Conflict of Interest or Certification R ole?" Journal o f Financial Economics, 40 (M arch 1996), pp. 373-402. Puri, Manju. "The Long-Term Default Perform ance of Bank Underw ritten Security Issues," Journal o f Banking and Finance, 18 (January 1994), pp. 397-418. Rajan, Raghuram G. "The Entry of Comm ercial Banks into the Securities Business: A Selective Survey of Theories and Evidence," in Universal Banking: Financial System Design Reconsidered. H om ew ood, IL: Irw in (1996), pp. 282-302. Rehm, Barbara A. "30 Banks Petition Fed to Increase 10% Cap on Securities A ctivities," American Banker (July 18,1994), p. 1. Saunders, Anthony. "Securities Activities of Comm ercial Banks: The Problem of Conflicts of Interest," Busi ness Review, Federal Reserve Bank of Philadelphia, July/A ugust 1985, pp. 17-27. Saunders, Anthony, and Ingo Walter. Universal Banking in the United States: What Could We Gain? What Could We Lose? New York: Oxford University Press, 1994. Walter, Ingo, ed. Deregulating Wall Street. New York: John Wiley and Sons, 1985. Digitized 18 FRASER for FEDERAL RESERVE BANK OF PHILADELPHIA Value at Risk: A New Methodology For Measuring Portfolio Risk Gregory P. Hopper* C V ^om m ercial banks, investment banks, insur ance companies, nonfinancial firms, and pen sion funds hold portfolios of assets that may include stocks, bonds, currencies, and deriva tives. Each institution needs to quantify the amount of risk its portfolio may incur in the course of a day, week, month, or year. For example, a bank needs to assess its po tential losses in order to set aside enough capi tal to cover them. Similarly, a company needs to track the value of its assets and any cash *Greg H opper is an econom ist in the Research Depart ment of the Philadelphia Fed. flows resulting from losses in its portfolio. An investment fund may want to understand po tential losses on its portfolio, not only to allo cate its assets better but also to fulfill its obliga tion to make set payments to investors. In ad dition, credit-rating and regulatory agencies must be able to assess likely losses on portfo lios as well, since they need to set capital re quirements and issue credit ratings. How can these institutions judge the likeli hood and magnitude of potential losses on their portfolios? A new methodology called value at risk (VAR or VaR) can be used to estimate these losses. This article describes the various meth ods used to calculate VAR, paying special at tention to VAR's weaknesses. 19 BUSINESS REVIEW WHAT IS VALUE AT RISK? Value at risk is an estimate of the largest loss that a portfolio is likely to suffer during all but truly exceptional periods. More precisely the VAR is the maximum loss that an institution can be confident it would lose a certain frac tion of the time over a particular period. Con sider a bank with a portfolio of assets that would like to characterize its potential losses using VAR. For example, the bank could specify a horizon of one day and set the frequency of maximum loss to 98 percent. In that case, a VAR calculation might reveal that the maximum loss is $1 million. Thus, on average, in 98 trading days out of 100, the loss on the portfolio will not exceed $1 million over a one-day horizon. But on two trading days in 100, losses will, on average, exceed $1 million. VAR can be used to assess the potential loss on a portfolio of assets generally. The user can specify any horizon and frequency of loss that fits his particular circum stances. But the method of calculating VAR depends not only on the horizon chosen but also on the kinds of assets in the portfolio. One method may yield good results with portfolios consisting of stocks, bonds, and currencies over a short ho rizon, but the same method may not work well over longer horizons such as a month or a year. If the portfolio contains derivatives, methods that differ from those used to analyze portfo lios of stocks, bonds, or currencies may be needed. VAR FOR A SINGLE SHARE OF STOCK Ultimately, we want to calculate VAR for a general portfolio of different assets, such as stocks, bonds, currencies, and options.1 Let's focus on the simplest case first: a single stock. A portfolio consisting of one asset will allow us to consider the different methods for assess- aAn option is a derivative security, i.e., its value is de rived from the value of some other asset. 20 JULY/AUGUST1996 ing VAR in a simple context. Then, we can gen eralize the discussion by considering how the calculation changes when the institution has a portfolio of many stocks, bonds, or currencies. Finally, we will consider how the inclusion of derivatives in the portfolio can dramatically change the methodology for calculating VAR. Randomness in the Stock Market. Let's con sider a portfolio consisting of a single share of stock worth $1 at the beginning of trading to day. We want to find the VAR over a one-day horizon at a 98 percent confidence level, that is, the largest one-day price drop we are likely to see on 98 out of every 100 trading days. Since VAR is essentially a statement about the likeli hood of losses on a stock, we need to character ize the unpredictability of daily changes in our stock's price. One way to picture the unpredictability of our stock's return over one day is to imagine the stock market spinning a roulette wheel. Of course, this is a fiction, but a useful one: econo mists have found that stock returns have a ran dom component. Suppose there are 100 equally likely out comes on the wheel, with each outcome corre sponding to a specific percentage daily price change or daily return for our stock.2 In gen eral, positive and negative returns are included on the wheel. To determine the return over one day, the stock market spins the roulette wheel. If the wheel comes up with a return of 25 per cent, our stock would be worth $1.25 at the end of the day. Alternatively, a spin of the wheel may generate a return of minus 25 percent, in which case our stock would be worth $0.75 at the end of the day. We can't say for sure what the daily return will be, but we know that it will be one of the outcomes on the wheel. 2In reality, w hen economists imagine stock returns on a wheel, they think of the wheel as having an infinite num ber of outcomes so that all possible returns are represented. To simplify the discussion, I have used 100 outcom es on the wheel as an approximation to an infinite-outcome wheel. FEDERAL RESERVE BANK OF PHILADELPHIA Value at Risk: A New Methodology for Measuring Portfolio Risk Gregory P. Hopper Finding the VAR for our $1 stock is particu larly simple if we know the returns on the rou lette wheel. Suppose we look at the outcomes on our roulette wheel and see that 98 of them involve returns bigger than minus 30 percent while two outcomes have returns smaller than minus 30 percent. Then we have found the VAR for our $1 stock: the VAR is $0.30 at a 98 per cent confidence level. We can be confident that 98 days out of 100 our daily stock loss will be no bigger than $0.30. But two days out of 100, the daily loss may indeed exceed $0.30. Summary Measures of Randomness. To find the VAR for our stock, we needed to know the 100 returns on the wheel. But how do we know what they are? Imagine that, every day, the market is spinning the wheel behind a cur tain. We can't see the outcomes on the wheel, but we do know which daily returns were se lected in the past—we can look them up in the newspaper. By categorizing past daily returns, we should be able to infer the outcomes on the wheel. For example, if we saw that daily returns of 10 percent occurred on five trading days in 100, on average, we can assume that five out comes on the wheel involve a 10 percent return. Similarly, if changes of minus 5 percent oc curred on 10 trading days in 100, on average, a return of minus 5 percent must correspond to 10 outcomes on the wheel. By continuing this analysis, we can associate price changes with all outcomes on the wheel. Then we will have reconstructed the wheel that the economy spins daily. Using our reconstructed wheel, we can easily find the VAR. A simpler way to do this reconstruction is to summarize the 100 returns on the wheel by using two numbers: the average return (mean) and the volatility (variance) of the returns. El ementary statistics teaches that if the returns follow a certain pattern, called the normal, or bell-shaped, distribution, all the outcomes on the wheel can be summarized by these two numbers. We can estimate the average return as an equally weighted average of past daily returns selected by the roulette wheel, returns that, again, could be looked up in the newspaper. For technical reasons, analysts often don't per form this calculation but assume instead that the average return is zero.3 The second num ber, the volatility, tells us how much the return is likely to deviate from its average value for any particular spin. The volatility, then, mea sures the capacity of the roulette wheel to gen erate extreme returns, whether positive or nega tive, with respect to the average value of zero. The higher the volatility of the roulette wheel, the more it tends to select large returns. We can estimate the volatility as an equally weighted average of past squared returns. We could use the same returns we looked up in the newspa per; we only need to square each change. Armed with the average return of zero and the volatility of our stock's returns, we can find the VAR over a one-day horizon at the 98 per cent confidence level by following a simple pro cedure. To calculate VAR for our stock, we need only multiply today's stock price of $1 times the square root of the volatility times a number corresponding to the 98 percent confidence level, called the confidence factor. The confi dence factor is derived from the properties of the normal distribution. At the 98 percent con fidence level, it equals 2.054.4 This procedure can be done on any day in 3Since the average return is estim ated very imprecisely, it m ay pay to set it to zero to avoid corrupting the rest of the VAR analysis. For more discussion on setting the aver age return equal to zero, see the article by Steven Figlewski and the 1995 article by David Hsieh. 4From elementary statistics, 2.054 standard deviations leave 2 percent of the norm al distribution in its left tail, which corresponds to stock losses occurring 2 percent of the time. If the confidence level were 95 percent, the confi dence factor would be 1.65, because 1.65 standard devia tions leave 5 percent of the norm al distribution in the left tail. 21 BUSINESS REVIEW JULY/AUGUST1996 the future as well. Let's assume that it's now tomorrow and the stock price is $0.95. If we wanted to calculate VAR, we would follow the same procedure as before but use a stock price of $0.95. We don't need to change the volatility or the confidence number: they don't vary from day to day. When VAR is calculated in this fash ion, we are using a constant volatility method. Time-Varying Volatility. The problem with the constant volatility method is that substan tial empirical evidence shows volatility is not constant from day to day but rather varies over time.5A look at a graph of the daily dollar re turn on the deutsche mark shows that volatil ity tends to cluster together (Figure 1). Notice that highly volatile times, characterized by large 5The evidence suggests that volatility is time-varying for short horizons such as up to a week or 10 days. For longer horizons, the evidence for time-varying volatility is weaker. If a firm is interested in calculating VAR over a much longer horizon, the tim e-varying volatility issue m ay not be so important. up-and-down swings in the exchange rate, tend to follow one another, while quiet periods, char acterized by smaller up-and-down swings, tend to follow each other as well. For example, vola tility seems to have been higher in 1991 than in 1990. A graph of the daily return on the S&P 500 confirms this impression for stock prices (Figure 2). The increase in volatility is particu larly apparent after the stock market crash in 1987. Time-varying volatility seems to be a gen eral feature of asset prices that is seen not only in currencies but also in stocks. Consequently, using the constant volatility method to calcu late VAR could be very misleading. What does time-varying volatility mean for our roulette wheel analogy? When the aver age return and the volatility don't vary from day to day, the returns on the wheel don't vary either. Thus, the market is spinning the same roulette wheel every day. But if the volatility is changing from day to day (time-varying vola tility), the returns on the wheel must also be changing; therefore the market is spinning a FIGURE 1 Daily Percent Dollar Return on Deutsche Mark Percent 22 FEDERAL RESERVE BANK OF PHILADELPHIA Value at Risk: A New Methodology for Measuring Portfolio Risk Gregory P. Hopper different wheel each day. If the market spins a different roulette wheel every day, VAR becomes more complicated. How do we know which returns will be on the wheel today? Equivalently, how do we know today's volatility? The most common solution to this problem was introduced in 1986 by economist Tim Bollerslev, who generalized work done by economist Robert Engle in 1982. Bollerslev's time-varying volatility technique, called the GARCH method, allows us to base our knowledge of today's roulette wheel on yesterday's wheel. Bollerslev's GARCH technique estimates the volatility of tod ay's roulette w heel using y esterd ay's estim ate of volatility and the squared value of y esterd ay 's return. If yesterday's return was large, in either a posi tive or negative direction, and yesterday's vola tility was high, today's roulette wheel will tend to have a high volatility. Thus, today's spin of the wheel will tend to produce large returns as well. In this way, large returns, positive or nega tive, would tend to follow one another, leading to periods of high and low volatility as we saw in Figures 1 and 2. How can we estimate today's volatility and find the VAR using B o llerslev 's GARCH method? The daily volatility using GARCH turns out to be a weighted average of past squared returns, just as it was in the constant volatility case. The difference is that the con stant volatility method weights past squared returns equally while Bollerslev's GARCH method weights recent squared returns more heavily than distant returns. It is easy to calculate volatility using the con stant volatility method. Bollerslev's GARCH method is much harder to implement: to find the right weight for each past squared return, we must employ a complicated, computer-intensive procedure. Once we have found today's volatility, we can multiply the confidence fac tor times the square root of today's volatility times today's stock price to find today's VAR. When we use Bollerslev's GARCH method, the FIGURE 2 Daily Percent Dollar Return on S&P500 Percent 10 6 2 -2 _______________1 ________1 _ _ _ _ _ _ L _ _ ll ■ _ _ _ _ ■_ _ l __ _ __ ________ _ _ ______________________ 1 _ _1 _ _____________________ o-i _____ L L i t ___ _________ - ___ * , T f n F W T ' lI W T ' H . U I i U U U liiJL J ----1 WT ------------ 1 TH n ^ -6 -10 -14 -18 -22 1 -2 6 1986 _ ! ________ 1 _______ 1 _______ _ _ 1987 1988 1989 1990 1 _______ 1 _______ _ _ 1991 1992 23 BUSINESS REVIEW JULY/AUGUST1996 confidence factor is the only number that does not change daily. RiskMetrics™. Bollerslev's GARCH method has found w idespread em pirical support among financial economists, but the difficulty in estimating daily volatilities has slowed its adoption by many institutions engaged in risk management. To make the calculations easier, J.P. Morgan introduced RiskMetrics™, a risk management system that includes techniques to approximate GARCH volatilities (see Pros and Cons o f Using RiskMetrics™ as a Risk-Man agement Tool). Like Bollerslev's method, the RiskMetrics™ estimate of daily volatility in volves a weighted average of past squared re turns, with recent squared returns weighted more heavily. The RiskMetrics™ weights are chosen to produce daily volatility estimates sim ilar to GARCH volatilities. The set of w eights calcu lated by the RiskM etrics™ method is easier to compute and can be used for any asset in the portfolio. For example, the analyst would use the same set of weights to calculate volatilities of stocks, bonds, and cur rencies. Bollerslev's GARCH method, in con trast, requires the computation of different weights for each volatility calculation, and each set of weights is harder to calculate than it would be using the RiskMetrics™ method.6 Other Methods. Two other methods of cal culating volatility are sometimes used. The first method relies on recognizing that pricing meth ods for options require the user to specify his estimate of the future volatility of an asset. For example, if a user wants to price an option on a stock using a method such as the popular BlackScholes method, he must specify an estimate of the volatility of the stock over the life of the option.7 Since option prices are observable in 6U nder the RiskMetrics™ m ethod, a different set of weights is calculated for each of a series of over 400 assets. The weights are then combined to yield a single composite set of weights that can be used for any asset in the portfo lio. Pros and Cons of Using RiskMetrics™ as a Risk-Management Tool Pros Cons • Com putationally convenient approxim ation to Bollerslev's GARCH method. Thus, will require relatively sm aller investm ent in research and inform ation systems. • Com m its user to a one-size-fits-all method: the GARCH method. This m ay be misleading for stocks, especially follow ing large changes in stock prices. GARCH may also not describe covariances well. • Not a proprietary system The methodology is explained in detail in J.P. M organ publications. • J.P. M organ publishes volatilities and correlations on a wide variety of assets free of charge. • Substantial third-party software support. 24 • There is no consensus on how well GARCH models forecast volatility. Even if GARCH models forecast volatility well in a statistical sense, that is, make small forecast errors, they may not forecast well in an econom ic sense. For example, the RiskM etrics™ volatility estimate m ay not m axim ize profits even if it does forecast volatility well in a statistical sense. • VAR may be the wrong m ethodology for the firm. FEDERAL RESERVE BANK OF PHILADELPHIA Value at Risk: A Neiv Methodology for Measuring Portfolio Risk Gregory P. Hopper the marketplace, the market's view of volatil ity can be backed out of the option price using the Black-Scholes formula. Volatility estimates inferred from option prices in this way are called implied volatilities. This method has two disadvantages that limit its appeal. First, options may not be traded on the particular asset of interest. Thus, implied volatility estimates may not be obtainable for some assets in the portfolio. Second, econo mists are unsure about whether implied vola tility estimates are better than GARCFf esti mates of daily volatility. The other method of estimating volatility is based on judgment. The user analyzes the eco nomic environment and forecasts volatility based on his subjective views. This method has limited appeal as well, since testing the valid ity of a subjective view is difficult. ing a portfolio whose volatility is lower than the volatility of each stock in the portfolio. Add ing more stocks to the portfolio would reduce the volatility further, provided the additional stocks' returns are not highly positively corre lated with the return of the initial portfolio. To account for this effect, we must also estimate the covariance between the stocks' returns. Once we know the stock returns' volatilities and covariances, we can calculate the volatility of the entire portfolio and find the VAR as before. As an example of the calculation, suppose we have invested $1 in stocks 1,2, and 3. Then by an elementary statistical formula, the daily volatility of the portfolio would be VAR FOR A PORTFOLIO OF ASSETS Up to this point, we have considered only how to calculate the VAR of a portfolio consist ing of a single stock. Now let's look at a portfo lio of two stocks. The principles we are about to discuss apply generally to portfolios of many assets, but we will consider just two stocks to make the ideas clear. As before, ultimately we want to find the volatility of the return on the portfolio. It's clear that the volatility of the portfolio should de pend on the volatility of the return of each stock in the portfolio. So, we need to estimate the volatilities of the returns of both stocks. But stock returns may covary as well. For example, if the covariance between the stocks in a port folio of two stocks is negative, then when one stock has a positive return, the other has a nega tive return, and vice versa. Thus, the two stocks dampen each other's swings in return, produc-7 * 7For an explanation of this method, see the article by Fischer Black and M yron Scholes. volatility (portfolio) = volatility(stock 1) + volatility(stock 2) + volatility(stock 3) + 2.0 x covariance(stock 1, stock 2) + 2.0 x covariance(stock 1, stock 3) + 2.0 x covariance(stock 2, stock 3) Notice that if the covariance between the daily returns of stocks 1,2, and 3 were zero, we could sum the volatilities of each stock to get the volatility of the portfolio. Thus, if covari ances between all assets were zero, we could find the VAR of each asset separately and then sum them to get the VAR of the portfolio. But since covariances are, in general, not zero, we can't, in general, find the VAR of individual assets and sum them to get the VAR of the port folio. Moreover, we can't find the VARs of as set classes such as stock and currency portfo lios separately and sum them. We must account for the covariances between asset classes as well. To calculate covariances between the assets' returns using the constant covariance method, we use an equally weighted average of the products of each stock's past daily returns. However, since economists have found evi dence that covariances change over time, it may be advisable to estimate time-varying covari ances using an exten sio n of B o llerslev 's 25 BUSINESS REVIEW GARCH method or the RiskMetrics™ GARCH approximation.8 JULY/AUGUST1996 often use an alternative method called Monte Carlo analysis. Using the volatility and covari ance estimates for the derivatives' underlying assets as well as a derivative pricing tool such as the Black-Scholes method, risk managers construct a new roulette wheel. The new wheel will still have 100 numbers, but each number will correspond to a potential change in the d e r iv a tiv e 's p rice . T h e c o m p u te r c a n th e n look at the largest loss the derivative will sustain for 98 of the outcomes. Let's suppose this loss is $0.01. Then the VAR of the derivative over a one-day horizon at the 98 percent confidence level is $0.01. Since RiskMetrics™ yields vola tility and covariance estimates, Monte Carlo evaluation of derivative portfolios can be done under J.R Morgan's system as well.9 * WHAT ABOUT DERIVATIVES? Many portfolios have significant numbers of derivatives such as futures, options, and swaps, all of which are securities whose value is de rived from the value of some other asset. Con sider a derivative on our $1 stock. We know how to find the VAR of the stock over a oneday horizon at the 98 percent confidence level: we find the volatility of its return and multiply its square root by the product of today's stock price and the confidence factor. But how can we find the VAR of a derivative on this stock? One method is to link the derivative to the underlying stock and use the standard VAR method. To do this, we use a derivative-pric ing method, such as the Black-Scholes model, to calculate a number called delta, which gives us a way to translate the derivative portfolio into the stock portfolio. A derivative's delta tells us how the derivative's price changes when the stock price changes a small amount. For ex ample, if the delta is 0.5, the derivative's price goes up half as much as the stock's price. For small price changes, a derivative with a delta of 0.5 behaves as if it is half a share of the $1 stock. So, using our estimate of the stock's vola tility, we could calculate VAR as before: by multiplying $0.50 times the square root of the stock's volatility times the confidence factor. A serious drawback to this method is that it works well only when stock price changes are small. For larger changes, delta itself can change dramatically, leading to inaccurate VAR esti mates. In general, we need to account for how delta changes, considerably complicating the analysis. To avoid this complication, risk managers WEAKNESSES OF VAR When properly used, VAR can give an insti tution an idea about the maximum losses it can expect to incur on its portfolio a certain frac tion of the time, making VAR an important riskmanagement tool. Using VAR calculations, an institution can judge how it should reallocate the assets in its portfolio to achieve the risk level it desires. But VAR methodology is not with out its weaknesses, and, improperly used, it may lead an institution to make poor risk-man agement decisions. This can happen for one of two reasons: either the VAR is incorrectly cal culated or the VAR is correctly calculated but irrelevant to the institution's real risk-manage ment goals. What Is the Best Method for Estimating Volatility? Bollerslev's GARCH method works better for currencies than it does for stock prices. Financial economists have found that stock volatility goes up more as a result of a large negative return than it does as a result of a large 8For further discussion on covariance GARCH tech niques, see the p aper by Robert Engle and Kenneth Kroner and the 1990 paper by Tim Bollerslev. 9For more detail on this process, see the RiskMetrics™ technical docum ent. For an example of a related method ology, see the 1993 articles by David Hsieh. 26 FEDERAL RESERVE BANK OF PHILADELPHIA Value at Risk: A New Methodology for Measuring Portfolio Risk Gregory P. Hopper positive return. A weakness of Bollerslev's GARCH method is that GARCH volatility esti mates don't depend on whether yesterday's return was positive or negative. Thus, this method can't allow for stock volatility's asym metric response to past returns. To account for this effect, financial econo mists have developed methods for estimating asymmetric volatilities.1 These methods are 0 important because they can give very different estimates of volatility for days following large stock returns than would the GARCH or RiskMetrics™ method. For small daily returns, Bollerslev's method, RiskMetrics™, and the asymmetric volatility method yield similar oneday-ahead volatility predictions, leading a user to think, perhaps, that one model is as good as the others for daily volatility predictions. But for large daily returns, the one-day-ahead vola tility predictions of these methods can be sub stantially different. If an asymmetric volatility method is appropriate for stock prices, both Bollerslev's method and RiskMetrics™ may understate one-day-ahead volatility whenever a large drop in stock prices occurred the previ ous day, thus producing a potentially substan tial underestimate of daily VAR. Similarly, the GARCH or RiskMetrics™ method could over estimate the VAR after a large increase in stock prices. Robert Engle and Victor Ng have provided evidence that a particular asymmetric volatil ity method well describes the volatility of Japa nese stock returns and that GARCH methods can substantially underpredict volatility follow ing large negative returns. Thus, VAR estimates of stock portfolios produced by GARCH or the RiskMetrics™ GARCH approximation should be viewed with caution if the calculations are done on days with large stock returns. Although having the right method for cal- culating the volatilities of assets is important, correctly calculating the covariances between the returns on assets is also important. Unfor tunately, not as much work has been done by financial econom ists to identify the right method for calculating covariances. To date, many methods have been proposed, but no consensus has yet emerged. Thus, we don't yet know for sure how we should handle covari ances in portfolios. This uncertainty introduces the risk that any method we use may substan tially under- or overestimate VAR. In particu lar, RiskMetrics™ commits the user to a special case of Bollerslev's GARCH method. Since we don't yet know whether Bollerslev's GARCH method is adequate in describing covariances, we should use even more caution in interpret ing results whenever we have used covariances in our VAR calculations. In the long run, the volatility estimates pro duced by GARCH methods tend, in general, to approach the values that the constant volatil ity method would have calculated. Thus, for horizons much longer than one day, using the constant volatility method to calculate VAR may be warranted.1 1 Frequency of Large Returns. Using either Bollerslev's GARCH model or the constant volatility method, we could find the VAR by assuming that the returns on the wheel follow a normal distribution. However, a substantial amount of evidence indicates that the normal distribution is inadequate because large daily returns, positive or negative, occur more often in the market than a normal distribution would suggest. One remedy is to use a different dis tribution for the price changes, one that gener ates more frequent large returns.1 Alternatively, 2 10The p ro to ty p ica l asy m m etric v o latility m odel is EGARCH. See the article by Daniel Nelson. 12For an example of this technique, see the article by Daniel Nelson. n See the article by David Hsieh (1993a) for a discussion about when the constant volatility m odel m ay be appro priate. 27 BUSINESS REVIEW we could use statistical methods that assume the returns follow the normal distribution, but which remain valid even if this assumption is mistaken. Whichever method we use, we are essen tially looking at the past frequencies and mag nitudes of returns and attempting to represent them on a reconstructed wheel. Even if we ac count for the nonnormality of returns during this process, there is still a problem: we're go ing to put on the wheel only those returns we saw in the past with the frequency we saw in the past. So, if some potential negative returns are rare or have not yet occurred, we may underrepresent them on the wheel, implying that the VAR will be underestimated. Structural Shifts in the Economy. VAR may be underestimated if the wheel the market is spinning suddenly changes in an unpredictable way because of a structural change in the un derlying economy. For example, consider the European Exchange Rate Mechanism (ERM), which kept daily returns of major European currencies small. In 1993, in response to eco nomic pressures, much larger returns were sud denly allowed. Thus, the volatility of the returns suddenly shot up faster than Bollerslev's GARCH method would have forecast based on past volatilities and returns. If we had calcu lated the VAR the day before the shift, we would have underestimated it because we would have used an estimate of the volatility that was too low. More subtly, since we never know when the economy may suddenly shift to higher or lower volatility as a result of a structural change, we will incorrectly estimate the VAR unless we explicitly account for this possibil ity. Because of the problems caused by infre quent large returns and structural shifts in the economy, it seems prudent, then, to supplement statistical calculations of VAR with judgmental estimates. For example, an institution could have asked its economists to project the likely price effects if the ERM suddenly allowed larger 28 JULY/AUGUST1996 price changes. These projections could be based on similar historical episodes, economic theory, and empirical experience. VAR estimates based on judgment could be generated for changes in central bank monetary regimes, political in stability, structural economic changes, and other events that have either never happened or happen infrequently. Liquidity of Assets. VAR m easures the maximum loss that an institution can expect a certain fraction of the time over a specific hori zon. Losses are measured by assuming that the assets can be sold at current market prices. However, if a firm has highly illiquid assets— meaning that they cannot quickly be resold— VAR may underestimate the true losses, since the assets may have to be sold at a discount. Credit Risk. Another potential problem for VAR is that the methods used to evaluate the assets in the portfolio may not properly treat credit risk. Suppose a bank buys a portfolio of derivatives from many different firms. The de rivatives are valuable to the bank because they impose obligations on the firms. For example, one of the derivatives may obligate a firm to sell foreign currency to the bank at a price be low the current market price, yielding a profit to the bank under some conditions, but it may also obligate the bank to deliver foreign ex change at a below-market price under other conditions. Using the Black-Scholes method and a Monte Carlo simulation, which assume no derivative credit risk, the bank calculates a VAR of $5 million at a 98 percent confidence rate for a three-month horizon. But if some of the firms may default on their obligations, the true value of these derivatives is lower than would be estim ated by the Black-Scholes method coupled with Monte Carlo analysis. Thus, the true value at risk is larger than $5 million. To account for this possibility when valuing derivatives, the bank should use a method that includes credit risk. For some ap plications, credit risk may be small enough to ignore, but, in general, users need to include FEDERAL RESERVE BANK OF PHILADELPHIA Value at Risk: A New Methodology for Measuring Portfolio Risk credit risk analysis in their VAR methods. Is VAR the Right Methodology? In many situations, VAR may not be the correct riskmanagement methodology. If we pick a specific loss such as $1 million, VAR allows us to esti mate how often we can expect to experience this particular loss. For example, using VAR we might estimate that we will lose at least $1 mil lion on one trading day in 20, on average. Dur ing some 20-day periods, we might lose less than $1 million. During other 20-day periods, we might lose more than $1 million on more than one day. VAR tells us how often we can expect to experience particular losses. It doesn't tell us how large those losses are likely to be. In particular, in any 20-day period, there is always one day on which the worst loss is experienced. If we want to know the size and frequency of the worst loss, VAR provides no guidance. One way of handling this is to use worstcase-scenario analysis (WCSA), proposed by Jacob Boudoukh, Matthew Richardson, and Robert Whitlelaw. WCSA might show that on the day with the worst price change in a 20day period, we can expect to lose at least $2.77 million 5 percent of the time, a number sub stantially bigger than $1 million. Thus, if a firm is interested in the size of a worst-case loss, VAR could underestimate it. CONCLUSION VAR is an important new concept in portfo lio risk management. It gives the maximum loss that an institution can expect to lose with a cer Gregory P. Hopper tain frequency over a specific horizon, and it can be calculated by using a constant volatility or time-varying volatility method. There are, however, problems in implementation and in terpretation. To implement VAR calculations, it is important to use the right method, espe cially under unusual circumstances such as stock market crashes. Although much progress has been made in describing how volatilities change through time, not as much progress has been made in the description of time-varying covariances. Thus, VAR numbers should be viewed with caution at this point. Besides the problem of identifying the right method, VAR measures may mislead unless they properly account for liquidity risk, rare or unique events, and credit risk. In many situa tions, it may not be the right risk-management concept. An institution may want to investigate an alternative, such as worst-case-scenario analysis. Despite the contribution that VAR can make to a firm's understanding of the risks in its port folio, these risks can be misunderstood if they are not communicated effectively to a manage ment that understands the value and limitations of sophisticated financial technology. Poor man agement practices, which could lead to unau thorized trades, may also contribute to this mis understanding. Thus, a firm should use VAR in the context of a broader risk-management culture, fostered not only by the firm's risk managers but also by its senior management. 29 BUSINESS REVIEW JULY/AUGUST1996 APPENDIX VAR and Capital Requirements for Market Risk In 1995, the Basle Com m ittee on Banking Supervision at the Bank for International Settlem ents (Basle Committee) issued a proposal for com m ent entitled "Internal M odel-Based Approach to M arket Risk Capi tal Requirem ents." This proposal would establish a VAR-based method of m easuring banks' portfolio risk. In January 1996, the Basle Com m ittee approved an approach that would allow banks to use their own internal risk-m anagem ent models or the Basle Com m ittee's standard model. The internal risk-m anagem ent models would be subject, however, to qualitative and quantitative restrictions. U.S. regulators are expected to im plement this approach for nine or 10 of the largest U.S. banks. Some exam ples of the restrictions the Basle Com m ittee would im pose on internal models are: Quantitative Criteria: • VAR m ust be com puted daily using a horizon of 10 trading days. • The confidence level should be set to 99 percent. • Models should account for changing delta when com puting VAR. In addition, VAR m odels should account for the im pact of tim e-varying volatility on option prices. • Banks m ay use covariances w ithin and across asset classes. Qualitative Criteria: • Banks should have independent risk-m anagem ent units that report directly to senior management. • VAR reports and analyses should be considered when setting trading limits. 30 FEDERAL RESERVE BANK OF PHILADELPHIA Value at Risk: A New Methodology for Measuring Portfolio Risk Gregory P. Hopper REFERENCES Academic Literature: Black, Fischer, and M yron Scholes. "The Pricing of Options and Corporate Liabilities," Journal o f Political Economy, 81 (1973), pp. 637-59. Bollerslev, Tim. "Generalized Autoregressive Conditional Fleteroskedasticity," Journal o f Econometrics, 31 (1986), pp. 307-27. Bollerslev, Tim. "M odelling the Coherence in Short-Run N om inal Exchange Rates: A M ultivariate General ized ARCH M odel," Review o f Economics and Statistics, 78 (1990), pp. 498-505. Engle, Robert F. "A utoregressive Conditional Heteroskedasticity with Estim ates of the Variance of U.K. Inflation," Econometrica, 50 (1982), pp. 987-1008. Engle, Robert F., and Victor K. Ng. "M easuring and Testing the Im pact of N ews on Volatility," Journal of Finance, 48 (1993), pp. 1749-78. Engle, Robert F., and Kenneth K. Kroner. "M ultivariate Sim ultaneous Generalized A rch," University of California at San Diego mim eo (1993). Figlewski, Steven. "Forecasting Volatility Using Historical D ata," New York University Working Paper, S-94-13 (1994). Hsieh, David A. "Im plications of Nonlinear Dynamics for Financial Risk M anagem ent," Journal o f Financial and Quantitative Analysis, 28 (1993a), pp. 41-64. Nelson, Daniel B. "C onditional Heteroskedasticity in Asset Returns: A New A pproach," Econometrica, 59 (1991), pp. 347-70. Practitioner Literature: Boudoukh, Jacob, M atthew Richardson, and Robert Whitelaw. "Expect the W orst," Risk (Septem ber 1995), pp. 100-01. Hsieh, David A. "A ssessing the M arket and Credit Risks of Long-Term Interest Rate and Foreign Currency Products," Financial Analysts Journal, July-August (1993b), pp. 75-79. Hsieh, David A. "N onlinear Dynam ics in Financial Markets: Evidence and Im plications," Financial Analysts Journal, July-A ugust (1995), pp. 55-62. RiskM etrics™ -Technical Docum ent, M organ Guaranty Trust Company, Global Research, New York, 1995. 31 FEDERAL RESERVE BANK OF PHILADELPHIA Business Review Ten Independence Mall, Philadelphia, PA 19106-1574
@FedFRASER