The U.S. Treasury securities market is a benchmark. As obligations of the U.S. government, Treasury securities are considered to be free of default risk. The market is therefore a benchmark for risk-free interest rates, which are used to forecast economic developments and to analyze securities in other markets that contain default risk. The Treasury market is also large and liquid, with active repurchase agreement (repo) and futures markets. These features make it a popular benchmark for pricing other fixed-income securities and for hedging positions taken in other markets.
The Treasury market's benchmark status, however, is now being called into question by the nation's improved fiscal situation. The U.S. government has run a budget surplus over the past two years, and surpluses are expected to continue (and to continue growing) for years. The debt held by the public is projected to fall accordingly and, under reasonable assumptions, much of the outstanding debt could be paid back within the next decade. The declining stock of debt may impact Treasury market liquidity and efficiency, thereby making Treasuries a less useful benchmark of risk-free interest rates as well as a less useful benchmark for pricing and hedging other fixed-income securities.
Moreover, recent market events have heightened concerns about the Treasury market's benchmark role and provided insight into how the market may perform in the future. For instance, yield spreads between Treasuries and other fixed-income securities widened sharply amid the financial markets crisis in the fall of 1998 in a so-called "flight to quality." A related "flight to liquidity" also caused yield spreads among Treasury securities of varying liquidity to widen sharply. Consequently, some of the attributes that make the Treasury market an attractive benchmark were adversely affected.
This paper examines the benchmark role of the U.S. Treasury market and the features that make it an attractive benchmark. In it, I examine the market's recent performance, including yield changes relative to other fixed-income markets, changes in liquidity, repo market developments, and the aforementioned flight to liquidity. I show that several of the attributes that make the U.S. Treasury market a useful benchmark were negatively affected by the events of fall 1998, and that some of these attributes did not quickly return to their precrisis levels. Furthermore, I demonstrate that the agency debt, corporate debt, and interest-rate swaps markets have features that might make them attractive benchmarks, and that the agency debt and swaps markets in particular are already assuming a limited benchmark role.
THE BENCHMARK U.S. TREASURY MARKET
A number of features contribute to the U.S. Treasury market's role as a benchmark. Treasuries are backed by the full faith and credit of the U.S. government and are therefore considered to be free of default risk.(1) Issuance to pay off maturing debt and raise needed cash has created a stock of Treasuries held by the public that totaled $3.6 trillion on September 30, 1999. The creditworthiness and supply of Treasury securities have resulted in a highly liquid round-the-clock secondary market with high levels of trading activity and narrow bid-ask spreads. Treasuries trade in an extremely active repo market in which market participants can borrow securities and finance their positions, as well as in an active futures market in which market participants can buy and sell securities for future delivery.
As Treasuries are considered to be free of default risk, yields on these securities represent risk-free rates of return. These risk-free rates are used in a variety of analytical applications to forecast interest rates, inflation, and economic activity. The rates are also used as benchmarks in the analysis and monitoring of other fixed-income and non-fixed-income securities. The performance of corporate bonds, for example, is often examined relative to that of Treasury securities, as the comparison allows one to separate yield changes due to changes in the risk-free rate from yield changes due to changes in credit risk (or due to the pricing of such credit risk).
Treasury securities are also used extensively for pricing securities and hedging positions in other U.S. dollar fixed-income markets. When a fixed-rate corporate debt issue is initially sold, for example, it is typically marketed in terms of a yield spread to a particular Treasury security rather than at an absolute yield or price.(2) Similarly, a position taken in a corporate debt issue is frequently hedged in the Treasury market. The ability to hedge in the Treasury market increases dealers' willingness to make markets and take positions in other markets, and thereby improves the liquidity of these other markets.
While the creditworthiness of Treasury securities is critical to their use as benchmark risk-free rates, the liquidity and efficiency of the market are also important. A highly liquid Treasury market ensures that observed Treasury prices are close to the market consensus of where prices should be and that changes in prices reflect revisions in the market consensus. An efficient market ensures that the risk-free rates implied by Treasury yields closely reflect the market's views of risk-free rates and that prices are no more than minimally affected by issue-specific differences in liquidity, supply, or demand.
When one evaluates the Treasury market's use as a benchmark for pricing and hedging purposes, features such as relative market performance, well-developed repo and futures markets, and liquidity are important. To be a good pricing or hedging vehicle, Treasury prices should be highly correlated with prices in other markets. A loss in a dealer's long position in mortgage-backed securities, for example, could then be offset by a dealer's short position in Treasuries. Hedges frequently involve taking short positions, so the ability to borrow Treasury securities at a low cost in the repo market is important. (The futures market can also be used to take short positions.) Finally, Treasury market liquidity is important, as hedgers must be able to buy and sell large Treasury positions quickly with minimal transaction costs.
Features of the Treasury market that make it a good benchmark thus depend on how one uses the market as a benchmark. Creditworthiness, liquidity, and efficiency are important as a reference benchmark for risk-free rates, but relative market performance is not important and active repo and futures markets are important only so far as they benefit liquidity. Relative market performance, active repo and futures markets, and liquidity are important as a pricing and hedging benchmark, but creditworthiness and efficiency are important only so far as they influence liquidity and relative market performance.
THE SHRINKING PUBLIC DEBT
As noted, the benchmark status of the U.S. Treasury market is being called into question by the country's improved fiscal situation. In fiscal year 1999, U.S. government revenues exceeded outlays by $123 billion, resulting in the first consecutive budget surpluses since 1956-57. As of July 1999, the U.S. Congressional Budget Office (1999), or CBO, was projecting growing budget surpluses for the next ten years (under existing laws and policies), rising from $161 billion in fiscal year 2000 to $413 billion in fiscal year 2009 (including Social Security trust funds).
The budget surpluses are reducing the stock of Treasury debt outstanding. Debt held by the public stood at $3.6 trillion on September 30, 1999, down from its peak of $3.8 trillion a year and a half earlier.(3) As of July 1999, the CBO was projecting that such debt would continue to fall over the next ten years, to $0.9 trillion at the end of fiscal year 2009. As a percentage of GDP, debt held by the public was projected to fall from 40.9 percent in 1999 to 6.4 percent in 2009.
The U.S. Treasury Department initially responded to its decreased funding needs by cutting issue sizes. In particular, bill sizes were cut sharply in March 1997 such that three-month bill sizes, for example, fell from the $11-$14 billion range to the $6.0-$8.5 billion range (excluding amounts issued to Federal Reserve Banks).
To continue to ensure large, liquid issues, the Treasury announced in May 1998 that it would limit further contraction of bill sizes and concentrate coupon offerings around larger, less frequent issues.(4) The Treasury thus reduced issuance of the five-year note from monthly to quarterly and eliminated issuance of the three-year note altogether. In August 1999, the Treasury announced that is was reducing the issuance frequency of the thirty-year bond from three times a year to twice a year and that it was considering reducing the issuance frequency of one-year bills and two-year notes.
To maintain large auction sizes and the liquidity of the most recent (on-the-run) issues, the Treasury proposed a debt buyback program in August 1999 and announced a revision to the original issue discount (OID) rules in November 1999. Under the buyback program, launched in January 2000, the Treasury will redeem outstanding unmatured Treasury securities by purchasing them from their current owners.(5) Changes to the OID rules allow the Treasury to reopen its most recent issues within one year of issuance without concern that the price of the issues may have fallen by more than a small amount.
Changes in policy or economic conditions may forestall a considerable shrinkage of the Treasury debt. Even if the market does shrink substantially, the Treasury Department's efforts to maintain large and liquid issues may stave off significant market repercussions. Nonetheless, the improved fiscal situation advances the possibility that the Treasury market will shrink considerably and that issuance sizes and/or frequencies will have to be reduced further.(6)
Reduced debt outstanding and reduced issuance sizes and/or frequencies would likely impact several Treasury market attributes. The market would likely become less liquid, with wider bid-ask spreads, reduced depth, and less trading activity. Reduced issuance sizes and/or frequencies would likely decrease the supply of lendable securities and thereby drive up the cost of borrowing issues in the repo market. Issue-specific differences in liquidity would probably become more important in determining prices. In turn, Treasuries might perform more disparately from other fixed-income securities. Persistent fiscal surpluses could thereby make the Treasury market a less attractive benchmark. While Treasuries will remain free of default risk, the reduced market liquidity and efficiency would decrease their usefulness as risk-free benchmarks. Greater costs of borrowing securities in the repo market combined with reduced liquidity and increasingly disparate performance would make Treasuries less desirable benchmarks for pricing securities or hedging positions in other markets.
THE RECENT PERFORMANCE OF THE BENCHMARK U.S. TREASURY MARKET
Recent financial market events have heightened concerns about the U.S. Treasury market's benchmark role and have provided direction as to how the market may perform in the future. In the fall of 1998, global financial market turmoil spurred investors to seek the safety of U.S. Treasury securities, driving prices up and yields down. As shown in Chart 1, the yield on the ten-year U.S. Treasury note dropped 125 basis points, to 4.16 percent, between August 19, 1998, and October 5, 1998. While this paper does not explain the events behind the financial crisis, a few notable events are included in the chart as reference points.(7)
One aspect of the financial crisis was a flight to quality in which yield spreads widened sharply between Treasuries and other fixed-income securities. Another aspect was a reduction in market liquidity, as an aversion to risk-taking decreased dealers' willingness to take positions and make markets. An increased cost of borrowing securities in the repo market also resulted from the financial crisis as did a sharp widening in yields between more and less actively traded Treasury securities.
This paper's analysis of these disruptions demonstrates why the benchmark topic is receiving increased attention and, more importantly, clarifies the market attributes that should be examined when evaluating alternative benchmarks. It also provides insight into how the Treasury market may perform if the outstanding debt starts declining more quickly, although it does not attribute the market's recent performance to the improved fiscal situation. Moreover, the analysis does not rate the Treasury market's performance as a benchmark, but rather illustrates the growing prominence of the benchmark topic and the features that are important to a benchmark market.
Relative Market Performance
The performance of Treasuries and other fixed-income securities diverged sharply in the fall of 1998. Investors sought the safety of risk-free Treasuries at the expense of securities with credit risk in the so-called flight to quality, driving a wedge between their performance. Chart 2 shows that yield spreads of various fixed-income securities over Treasuries widened between mid-August and mid-October 1998, and remained fairly wide afterward. The yield spread between investment-grade corporate debt securities and Treasuries, for example, widened from 74 basis points on August 13, 1998, to 128 basis points on October 19, 1998. It was 116 basis points on October 31, 1999.
The widening of the spread in the fall of 1998 is not unprecedented. Credit spreads often rise during or preceding a recession, and they were quite high in the early 1980s, for example. One of the attractive features of Treasury securities is their absence of default risk. This means that Treasury yield changes do not reflect changes in credit risk, by definition, and that Treasuries are inherently limited in their ability to serve as good hedges of fixed-income securities that contain credit risk.
Despite the widening of the spread, there does not seem to have been a fundamental shift in the relationship between Treasury yield changes and other fixed-income yield changes. An analysis of weekly yield changes shows that Treasuries remained highly correlated with other fixed-income securities during the height of the financial crisis (Table 1). The correlation between ten-year Treasury yield changes and investment-grade corporate yield changes, for example, fell only slightly--from 0.975 before the crisis to 0.965 during the crisis and to 0.963 after the crisis.(8, 9)
Correlations of U.S. Treasury and Other Fixed-Income Yield Changes
Investment-Grade Mortgage-Backed Period Corporate Security Precrisis: July 3, 1997-Aug. 14, 1998 0.975 0.956 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.965 0.957 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.963 0.924 Full sample: July 3, 1997-Oct. 29, 1999 0.966 0.945 Fannie Mae High-Yield Period Benchmark Swap Corporate Precrisis: July 3, 1997-Aug. 14, 1998 0.976 0.987 0.473 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.970 0.968 0.199 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.956 0.961 0.429 Full sample: July 3, 1997-Oct. 29, 1999 0.964 0.970 0.286
Source: Author's calculations, based on data from Bloomberg, Goldman Sachs, and Merrill Lynch.
Notes: The table reports the correlations of weekly yield changes between the on-the-run ten-year U.S. Treasury note and the indicated index or security. Correlations with the Fannie Mae benchmark are limited to the period starting February 3, 1998. The investment-grade corporate yield is the industrials ten-year A2/A yield from Bloomberg. The mortgage-backed security yield is a weighted-average, option-adjusted yield calculated by Goldman Sachs. The Fannie Mae benchmark yield is the on-the-run ten-year benchmark note yield from Merrill Lynch, via Bloomberg. The swap rate is the ten-year semiannual fixed rate versus three-month LIBOR compiled by Bloomberg from various sources. The high-yield corporate yield is from Merrill Lynch's High-Yield Master Index, via Bloomberg.
The disparate performance of Treasury securities and other fixed-income securities raises questions about the attractiveness of Treasuries as hedging vehicles. Those who shorted Treasuries as a hedge preceding the widening of the spread in the fall of 1998 found that their losses on Treasuries more than offset any gains they may have had on their long positions. Nonetheless, the widening of the spread was not unprecedented, and Treasury yield changes maintained a high correlation with other fixed-income yield changes.
While the Treasury market was seen as a safe and liquid haven for investors in fall 1998, its liquidity was adversely affected nonetheless. One measure of liquidity is the bid-ask spread, or the difference between quoted bid and offer prices. As shown in Chart 3, spreads in the interdealer Treasury market widened sharply in fall 1998 for the on-the-run ten-year note and had not returned to precrisis levels as of October 1999. The ten-year note typically trades with a spread of 1/64 or 1/32 of a point (where one point equals 1 percent of par), but it traded with nearly a 3/32 average spread on October 9, 1998, and just over a 1/32 spread on October 29, 1999. For the ten-year note, 1/32 of a point equals just under half a basis point in yield terms.
Another measure of liquidity is the depth of the market. Market depth refers to the quantity of securities that dealers are willing to buy and sell at various prices, and is measured here by the average quantity firmly offered at the best quoted bid and offer prices in the interdealer market. As shown in Chart 4, the quoted depth of the on-the-run ten-year note fell from the $9-$11 million range in July and August 1998 to roughly $6 million in October 1998. Quoted depths did not recover quickly after fall 1998, averaging slightly more than $5.5 million in 1999 (through October).
One other measure of liquidity is trading volume. Volume is not an ideal measure of liquidity, as it may reflect dealers' eagerness to rebalance and hedge positions amid market turmoil, rather than their willingness to take positions and make markets. In fact, the volume numbers in Chart 5 show that trading activity actually increased throughout August and into early September 1998. Trading activity then declined fairly steadily throughout the fall before dropping off sharply at the end of the year; it remained lower than usual through October 1999.
The evidence suggests that Treasury market liquidity was adversely affected by the events of fall 1998 and that it did not recover quickly. While the market was quite volatile in fall 1998--and somewhat more volatile after the crisis than before it--such volatility does not explain the diminished liquidity.(10) The events of fall 1998, concerns about Y2K, the withdrawal of market participants, and the reluctance of remaining participants to take risks are some of the factors that may have inhibited market liquidity even after the crisis.
Reduced market liquidity can diminish the attractiveness of the Treasury market both as a risk-free benchmark and as a benchmark for pricing and hedging. Decreased liquidity increases the chances that implied risk-free rates will deviate from the market consensus as to where risk-free rates should be. Decreased liquidity also raises hedgers' direct costs of trading and reduces their ability to take or unload large positions quickly with minimal price impact. Despite the disruptions to Treasury market liquidity, it should be noted that the market remains highly liquid and that it may have been less disrupted by liquidity problems in fall 1998 than were other fixed-income markets.
The Repo Market
A repo is an agreement to exchange collateral for cash with a simultaneous agreement to buy back the collateral at a specified price at some point in the future. A dealer owning a particular Treasury note, for example, might agree to sell that security to another dealer while simultaneously agreeing to buy back the security the next day. The first dealer can thus use the repo market to finance its positions, often at a favorable rate, while the second dealer can use the repo market to borrow and then sell securities it does not hold in its portfolio.
The repo market for Treasury securities was temporarily disrupted by the events of fall 1998. One measure of disruption examines the spread between the general collateral rate and the collateral rate on a particular security. When an issue is in high demand, a dealer in effect lends funds at a rate below the rate that would otherwise be required to borrow a security, and the issue is said to be "on special." Table 2 shows that the on-the-run two-year note, five-year note, and thirty-year bond (but not the ten-year note) traded at an increased rate of specialness during the fall 1998 crisis, but that specialness declined after the crisis. The five-year note, for example, was lent at an average overnight rate that was 77 basis points below the general collateral rate before the crisis, 126 basis points during the crisis, and 75 basis points after the crisis.
Rope Specialness of On-the-Run U.S. Treasury Coupon Securities Basis Points
Two- Five- Ten- Thirty- Period Year Year Year Year Precrisis: July 1, 1997-Aug. 14, 1998 21.0 76.9 165.8 120.6 (30.4) (80.5) (135.8) (135.0) Crisis: Aug. 17, 1998-Nov. 20, 1998 52.8 126.1 115.6 211.1 (86.6) (149.3) (143.4) (164.9) Postcrisis: Nov. 23, 1998-Oct. 29, 1999 35.3 75.0 200.3 120.1 (48.6) (86.2) (155.0) (123.9) Full sample: July 1, 1997-Oct. 29, 1999 30.4 81.8 173.9 130.8 (48.5) (94.3) (146.8) (137.4)
Source: Author's calculations, based on data from GovPX.
Note: The table reports the means and standard deviations (in parentheses) of the daily average differences between the overnight general collateral rate and the collateral rates on the indicated on-the-run securities.
Repo activity in on-the-run coupon securities was not negatively affected by the events of fall 1998. As shown in Table 3, overnight repo trading volume increased in fall 1998 for the two-year and five-year notes, but it fell for the ten-year note and thirty-year bond. After fall 1998, repo activity changed little for the two-year and five-year notes, but it increased for the ten-year note and thirty-year bond. Overall, repo activity was higher after the crisis than it was before it for three of these four securities (all but the ten-year note). Repo trading volume numbers do not suggest that the use of Treasuries as hedging vehicles declined as a result of the fall 1998 crisis.
Repo Trading Volume of On-the-Run U.S. Treasury Coupon Securities
Billions of U.S. Dollars
Two- Five- Ten- Thirty- Period Year Year Year Year Precrisis: July 1, 1997-Aug. 14, 1998 5.69 7.42 10.39 4.09 (2.94) (3.09) (4.00) (2.10) Crisis: Aug. 17, 1998-Nov. 20, 1998 8.33 8.72 8.44 3.54 (3.50) (3.14) (2.79) (1.69) Postcrisis: Nov. 23, 1998-Oct. 29, 1999 8.31 8.78 9.54 4.25 (3.15) (3.19) (4.61) (1.87) Full sample: July 1, 1997-Oct. 29, 1999 7.04 8.11 9.82 4.09 (3.36) (3.20) (4.18) (1.97)
Source: Author's calculations, based on data from GovPX.
Note: The table reports the means and standard deviations (in parentheses) of daily overnight repurchase agreement trading volume in the indicated on-the-run securities as reported to GovPX.
Increased repo market specialness can decrease the attractiveness of Treasury securities as hedging vehicles because it makes borrowing securities more costly. Increased borrowing costs may also reduce market liquidity, further hurting the attractiveness of the Treasury market for various purposes, including pricing, hedging, and as a benchmark of risk-free rates. The evidence suggests, however, that the cost of borrowing on-the-run Treasury securities increased only briefly during the fall 1998 crisis and that repo market activity generally did not decline either during or after fall 1998.
One of the most striking developments in fall 1998 was a divergence in performance between more and less actively traded Treasury securities. As shown in Chart 6, the yield spread between the on-the-run five-year note and a comparable off-the-run security rose sharply in late August 1998 and again in mid-October 1998, reaching 25 basis points on October 15, 1998.(11) Table 4 shows that the comparable spread also widened sharply in fall 1998 for the two-year note and the thirty-year bond, albeit not for the ten-year note. On-the-run Treasuries generally became relatively more valuable as investors sought not only the safety of Treasury securities but also the liquidity of the on-the-run issues in the so-called flight to liquidity.(12) After the crisis, spreads remained high on the five-year note and the thirty-year bond, they increased for the ten-year note, but they declined for the two-year note.
Off-the-Run/On-the-Run Yield Spreads of U.S. Treasury Coupon Securities
Two- Five- Ten- Thirty- Period Year Year Year Year Precrisis: July 1, 1997-Aug. 14, 1998 2.80 4.48 7.87 5.01 (1.80) (1.90) (1.71) (1.71) Crisis: Aug. 17, 1998-Nov. 20, 1998 11.62 16.68 6.63 12.99 (5.76) (4.89) (3.30) (4.65) Postcrisis: Nov. 23, 1998-Oct. 29, 1999 5.02 17.93 13.55 13.50 (2.37) (2.75) (6.93) (1.83) Full sample: July 1, 1997-Oct. 29, 1999 4.72 11.33 10.03 9.36 (3.86) (7.14) (5.54) (4.78)
Source: Author's calculations, based on data from Bear Stearns and GovPX.
Note: The table reports the means and standard deviations (in parentheses) of the daily off-the-run/on-the-run yield spreads of the indicated securities. The spreads are calculated as the predicted yields less the market yields, where the predicted yields are those of comparable-duration off-the-run securities as derived from a model of the yield curve estimated with off-the-run prices.
Another development in fall 1998 was a divergence in pricing among off-the-run securities, possibly due to a decline in Treasury market arbitrage. The efficiency of the Treasury market typically results in off-the-run securities of similar maturity trading relatively close to one another in terms of yield. When Treasury yields are plotted against time to maturity, they usually form a relatively smooth curve, as shown for May 13, 1998 (Chart 7). The smoothness of the yield curve over time is estimated here as the median absolute error between market yields and the yields predicted by a term structure model.(13) As shown in Chart 8, the median rose sharply between late August and mid-October 1998--peaking at 2.3 basis points on October 8, 1998--and remained relatively high after the crisis.
The relative performance of Treasuries in the fall of 1998 is summarized in Chart 9, which plots yields against years to maturity for October 9, 1998. The chart shows the wide dispersion of off-the-run yields, as documented in Chart 8. It also shows the wide yield spreads between on-the-run coupon securities and comparable-maturity off-the-run securities, as shown in Chart 6.
The divergent performance of Treasury securities raises concerns about the market's usefulness both as a risk-free interest-rate benchmark and as a benchmark for pricing and hedging. Differences in the liquidity or specialness of Treasury securities can result in different implied risk-free rates, raising the issue of which risk-free rate is the appropriate one. Such differences also create an additional performance wedge between Treasuries and other fixed-income securities, possibly decreasing their correlation.(14) Nevertheless, while the divergent performance of Treasuries may hinder their role as a benchmark, it is noteworthy that this divergence may largely reflect market participants' demand for the securities' safety and liquidity. Characteristics that make the Treasury market an attractive benchmark in some ways may therefore result in performance undesirable of a benchmark in other ways.
The recent performance of the benchmark U.S. Treasury market and the improved fiscal situation raise the issue of which market or markets might serve as a future benchmark. While there is no obvious U.S. dollar alternative for risk-free rates, several markets are already assuming a limited benchmark role for pricing and hedging securities and as reference rates for monitoring and analytical purposes. These markets include the agency debt market, the corporate debt market, and the interest-rate swaps market. Each is examined in turn with regard to the features that make a good benchmark market.
The Agency Debt Market
Agency securities are obligations of federal government agencies or government-sponsored enterprises such as Fannie Mae, Freddie Mac, the Federal Home Loan Banks (FHLBanks), the Farm Credit Banks, Sallie Mae, and the Tennessee Valley Authority.(15) The agencies issue debt securities to finance activities that are supported by public policy, including home ownership, farming, and education. The securities typically are not backed by the full faith and credit of the U.S. government, as is the case with Treasury securities, and therefore trade with some credit risk. They are nevertheless considered to be of very high credit quality and are rated Aaa/AAA by the major rating agencies.
Seeking to capitalize on the market's interest in large, liquid issues amid reduced Treasury supply, the agencies have introduced their own benchmark debt issuance programs, starting with Fannie Mae's Benchmark Notes Program in January 1998. The programs provide for the regular issuance of large-size, noncallable coupon securities in a range of maturities (originally two to ten years), and thus mimic the Treasury Department's issuance practices. The benchmark securities are intended to appeal to investors who might typically buy Treasury securities, and are promoted as Treasury substitutes.(16)
The agency benchmark programs have expanded rapidly in their breadth and depth. Freddie Mac introduced its Reference Notes Program in April 1998; the FHLBanks introduced their Tap Issuance Program in July 1999 and also increased issuance sizes in their Global Debt Program; and the Farm Credit Banks introduced their Designated Bonds Program in March 1999. The programs have expanded beyond their original scope with the introduction of callable benchmark programs, the issuance of longer term securities, and the announcements of auction schedules. In November 1999, both Fannie Mae and Freddie Mac announced the introduction of benchmark bill programs, with weekly auctions of large-size discount securities.
As shown in Table 5, benchmark issues of the three largest agencies generally range from $3-$6 billion in size (as of October 1999), and thus are about one-fifth to one-half as large as comparable Treasury issues. As shown in Table 6, total benchmark issuances in 1999 through October were roughly $40 billion for each of the three largest agencies, versus $234 billion in Treasury coupon security issuances. Agency benchmark debt outstanding is even smaller relative to that of the Treasury Department, due to the recent introduction of the agency benchmark programs. Fannie Mae, for example, had $94 billion in noncallable benchmark securities outstanding on October 31, 1999 (Fannie Mae 1999b), whereas the Treasury Department had $2.4 trillion in marketable fixed-rate coupon securities outstanding (Bureau of the Public Debt 1999).
Issue Sizes of Agency and U.S. Treasury Coupon Securities as of October 31,1999 Billions of U.S. Dollars
Fannie Mae Freddie Mac FHLBanks Issue Benchmark Reference Global Two-year -- 5.0(a) 3.0 Three-year 3.0 5.0 3.0 Five-year 6.5(a) 3.0 -- Seven-year -- -- -- Ten-year 3.5 6.0 -- Thirty-year 4.25(a) -- -- FHLBanks U.S. Issue Tap Treasury Two-year 3.5(a) 15.0 Three-year 3.4(a) -- Five-year 2.0(a) 15.0 Seven-year 1.1(a) -- Ten-year 0.6(a) 12.0 Thirty-year -- 10.0
Sources: Bloomberg; FHLBanks, Office of Finance; Freddie Mac.
Notes: The table reports the sizes of the most recent noncallable benchmark coupon issues as of October 31, 1999. Securities more than one year old are excluded. FHLBanks Global Debt Program issues exclude a $1 billion one-year coupon issue and a $3.5 billion issue originally issued with three years to maturity. U.S. Treasury issue sizes exclude amounts issued to refund maturing securities of Federal Reserve Banks as well as amounts bid for by Federal Reserve Banks on behalf of foreign and international monetary authorities.
Issuance of Agency and U.S. Treasury Coupon Securities from January to October 1999 Billions of U.S. Dollars
Fannie Mae Freddie Mac FHLBanks Issue Benchmark Reference Global Two-year -- 9.0 17.0 Three-year 3.0 10.5 9.0 Five-year 19.5 9.0 -- Seven-year -- -- -- Ten-year 15.5 13.0 -- Thirty-year 4.25 -- -- Total 42.25 41.5 26.0 FHLBanks U.S. Issue Tap Treasury Two-year 3.9 135.0 Three-year 3.4 -- Five-year 2.3 45.0 Seven-year 1.3 -- Ten-year 0.7 34.0 Thirty-year -- 20.0 Total 11.7 234.0
Sources: Bloomberg; Fannie Mae; FHLBanks, Office of Finance; Freddie Mac.
Notes: The table reports noncallable benchmark coupon security issuance between January 1 and October 31, 1999. The FHLBanks Global Debt Program two-year amount includes a one-year issue as well as the reopenings of an old three-year note at two-and-a-half and two-and-a-quarter years to maturity. The FHLBanks two-year Tap Issuance Program amount includes a (me-and-a-half-year issue, the three-year amount includes a two-and-a-half-year issue, the five-year amount includes a four-year issue, the seven-year amount includes an eight-year issue, and the ten-year amount includes a fifteen-year issue. U.S. Treasury issuance excludes amounts issued to refund maturing securities of Federal Reserve Banks as well as amounts bid for by Federal Reserve Banks on behalf of foreign and international monetary authorities.
The stock of agency debt securities outstanding provides a guide as to how large the agency benchmark programs can become. As of June 30, 1999, agency debt outstanding totaled $1.4 trillion, versus $3.7 trillion of Treasury debt held by the public (Federal Reserve Bulletin 1999; Treasury Bulletin 1999). As shown in Chart 10, the agency debt market has grown rapidly in recent years, whereas the Treasury market has leveled off. Even if agency debt growth slowed to the rate of GDP growth (projected by the CBO), the agency debt market would surpass the U.S. Treasury market in size in fiscal year 2007 if the Treasury market shrinks according to the CBO's July 1999 projections.
The performance of agency securities versus other fixed-income securities suggests that agencies may be good pricing and hedging benchmarks. Fixed-income securities with credit risk (or spread products) largely moved together during and after the fall 1998 crisis, as shown in Chart 2. The correlations of the weekly yield changes of the Fannie Mae ten-year benchmark with those of other spread products are high, as shown in Table 7, and are comparable to those of Treasuries with other spread products (Table 1). The correlation between the Fannie Mae benchmark and mortgage-backed securities, for example, was 0.954 for the postcrisis period, versus 0.924 for Treasuries and mortgage-backed securities.
Correlations of Fannie Mae Benchmark and Other Fixed-Income Yield Changes
Investment-Grade Period Corporate Precrisis: Feb. 3, 1998-Aug. 14, 1998 0.948 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.915 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.934 Full sample: Feb. 3, 1998-Oct. 29, 1999 0.926 Mortgage-Backed Period Security Swap Precrisis: Feb. 3, 1998-Aug. 14, 1998 0.926 0.976 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.949 0.990 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.954 0.983 Full sample: Feb. 3, 1998-Oct. 29, 1999 0.950 0.985 High-Yield Period Corporate U.S. Treasury Precrisis: Feb. 3, 1998-Aug. 14, 1998 0.422 0.976 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.292 0.970 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.450 0.956 Full sample: Feb. 3, 1998-Oct. 29, 1999 0.309 0.964
Source: Author's calculations, based on data from Bloomberg, Goldman Sachs, and Merrill Lynch.
Notes: The table reports the correlations of weekly yield changes between the on-the-run ten-year Fannie Mae benchmark note and the indicated index or security. The Fannie Mae benchmark yield is from Merrill Lynch, via Bloomberg. The investment-grade corporate yield is the industrials ten-year A2/A yield from Bloomberg. The mortgage-backed security yield is a weighted-average, option-adjusted yield calculated by Goldman Sachs. The swap rate is the ten-year semiannual fixed rate versus three-month LIBOR compiled by Bloomberg from various sources. The high-yield corporate yield is from Merrill Lynch's High-Yield Master Index, via Bloomberg.
Agency market liquidity does not yet approach that of the U.S. Treasury market. As shown in Table 8, trading in agency coupon securities by the primary government securities dealers averaged $7.9 billion per day before the fall 1998 crisis, versus $183 billion in Treasury coupon securities. Trading in agency coupons increased to $10.7 billion per day after the crisis, while comparable Treasury trading fell, but agency coupon trading still equaled only 6.8 percent of postcrisis Treasury trading. Fannie Mae reports that its benchmark securities have liquidity comparable to off-the-run Treasury securities, with bid-ask spreads of 0.5 to 2.0 basis points (Fannie Mae 1999a).
Agency and U.S. Treasury Coupon Security Trading Volume
Agency U.S. Treasury Securities Securities (Billions of (Billions of Period U.S. Dollars) U.S. Dollars) Precrisis: Jan. 22, 1998-Aug. 12, 1998 7.9 183.3 (1.3) (29.1) Crisis: Aug. 13, 1998-Nov. 18, 1998 9.5 223.1 (1.2) (34.1) Postcrisis: Nov. 19, 1998-Oct. 27, 1999 10.7 156.7 (3.0) (25.8) Full sample: Jan. 22, 1998-Oct. 27, 1999 9.6 175.2 (2.7) (36.5) Agency- U.S. Treasury Ratio Period (Percent) Precrisis: Jan. 22, 1998-Aug. 12, 1998 4.4 (0.8) Crisis: Aug. 13, 1998-Nov. 18, 1998 4.3 (0.6) Postcrisis: Nov. 19, 1998-Oct. 27, 1999 6.8 (1.8) Full sample: Jan. 22, 1998-Oct. 27, 1999 5.7 (1.9)
Source: Author's calculations, based on data from the Federal Reserve Bank of New York and Federal Reserve Bulletin (1999).
Notes: The table reports the means and standard deviations (in parentheses) of average daily coupon security trading volume (reported weekly) of the primary government securities dealers.
An active overnight repo market in agency securities has developed, allowing market participants to borrow securities for hedging and trading purposes, although an active term repo market has not yet emerged. Agency issues sometimes trade on special, although typically still close to general collateral. As a result, Fannie Mae reports that its issues are largely unaffected by issue-specific differences in specialness or liquidity (Fannie Mae 1999a). Unlike the Treasury market, there is no futures market for agency securities.(17)
Agency debt securities are treated as benchmarks in a few respects. First, the yields on benchmark securities are used as barometers of the agency market for monitoring and analytical purposes. Second, agencies are used as hedging vehicles to a certain extent, particularly for mortgage-backed securities. Finally, at least one new debt issue has been priced relative to a benchmark agency security as of October 1999.(18)
Several attributes favor the agency debt securities market as a benchmark market. Namely, the performance of agency securities is highly correlated with that of other spread products, and agencies--because of their credit risk--have the potential to be better pricing and hedging vehicles than Treasuries. The market is also reasonably liquid, agencies trade in an active overnight repo market, and agencies reportedly have been relatively unaffected by issue-specific differences in liquidity or specialness. Steps taken by the agencies to increase issuance sizes are likely to improve market liquidity, and the announcements of issuance schedules and the resulting predictability of agency issuance are likely to improve activity in the term repo market.(19)
Nevertheless, other attributes do not favor the agency debt securities market as a benchmark. Credit risk, for example, may cause agencies to trade in line with other spread products, but the presence of such risk also means that there is an idiosyncratic risk component to agency securities that could become important in the future. Market liquidity also does not compare with that of the Treasury market, the overnight repo market is less active than the Treasury market, the term repo market is not active at all, and there is not yet an agency futures market. Furthermore, while agency securities may not be affected by issue-specific differences in liquidity or repo market specialness, this condition may reflect the lack of demand among market participants to borrow and trade agency benchmark issues. If the popularity of agency benchmark securities increases, issue-specific differences may become more important.
The Corporate Debt Market
Corporate debt securities are issued to meet a variety of longer term corporate financing needs. Their credit risk varies significantly across issues, from relatively safe Aaa/AAA-rated issues to non-investment-grade Ba/B, B/B, and Caa/CCC issues. The corporate debt market is larger, but far more segmented, than the agency debt market, with debt outstanding totaling $2.9 trillion on September 30, 1999 (Bond Market Association 1999).
Corporate issuers recently have increased issuance sizes and regularity to appeal to investor demand for large, liquid issues. Ford Motor Company, in particular, announced its Global Landmark Securities (GlobLS) Program in June 1999, modeled on the programs of Fannie Mae and Freddie Mac. Under the program, Ford and its financing subsidiary, Ford Motor Credit Company, announced that they would bring offerings of at least $3 billion to market two to four times per year. Ford issued $8.6 billion in four parts in July 1999 as part of the program and $5 billion in a single part in October 1999.
While the Ford issuances are large by corporate standards, they are significantly less than those of the agencies. In 1999, through October, Ford issued $13.6 billion in its GlobLS Program, as opposed to roughly $40 billion each in the three largest agencies' benchmark programs. It is worth noting that Ford's issuances are constrained by the size of the company's balance sheet. Ford had debt outstanding of $144 billion on June 30, 1999 (Ford Motor Company 1999), versus $500 billion for Fannie Mae, $437 billion for the FHLBanks, and $314 billion for Freddie Mac (Federal Reserve Bulletin 1999).
Liquidity of the large Ford issues is reportedly favorable, with bid-ask spreads of 1 to 2 basis points, compared with 3 to 5 basis points for smaller issues of similar quality (Bloomberg 1999). There is no futures market for Ford or other corporate issues, and corporate issues are not actively traded in the repo market.
Ford GlobLS play a limited benchmark role in the corporate market. They are used as reference rates for monitoring the performance of the corporate market, for evaluating other outstanding corporate debt securities, and for helping to decide how other new corporate debt issues should be priced. Hedging activity using corporate issues is limited.
The corporate market's potential as a benchmark is limited by its fragmented nature, with the largest corporate issuers being smaller than the large agency issuers. Corporates also do not have the creditworthiness of the agencies (Ford is rated Al/A), so that firm-specific developments may be more important in explaining the performance of any particular issuer's securities. Nevertheless, the trend toward increased issuance sizes and regularity will likely increase the role of corporates as benchmarks for monitoring and analysis within the corporate market.
The Interest-Rate Swaps Market
An interest-rate swap is an agreement between two parties to exchange one stream of interest payments for another stream. The most common interest-rate swap is used to exchange fixed interest-rate payments for floating interest-rate payments for a given principal amount and period of time. The floating rate in such contracts is often based on the London Interbank Offer Rate (LIBOR)--the rate that banks charge one another for funds in the Eurodollar market.
Swap rates are quoted in terms of the fixed rate that must be paid to convert to a floating rate. At the close of September 30, 1999, for example, the quoted ten-year swap rate on Bloomberg was 6.85 percent. An entity therefore had to make semiannual fixed interest payments for ten years at an annual rate of 6.85 percent to get semiannual floating interest payments for ten years based on three-month LIBOR (for the same principal amount). Swap rates are often quoted relative to the Treasury benchmark, so that the ten-year spread on September 30 was quoted as 97 basis points (calculated as the 6.85 percent swap rate less the 5.88 percent yield on the on-the-run ten-year Treasury note). Swap rates exceed those on Treasuries mainly because the floating payments are based on a rate that contains credit risk (LIBOR is an Aa/AA rate).
Since they are based on a floating rate that contains credit risk, swap rates often change in line with yields on other spread products. Swap spreads thus widened sharply in fall 1998 along with those of corporates, agencies, and mortgage-backed securities, as shown in Chart 2. Correlations of weekly changes in ten-year swap rates with yields of other spread products, shown in Table 9, are close to those of Treasuries with other spread products (Table 1). The correlation with Fannie Mae's benchmark note, for example, is 0.985 for swaps, versus 0.964 for Treasuries (for the full sample period).
Correlations of Swap Rate and Other Fixed-Income Yield Changes
Investment- Mortgage- Grade Backed Period Corporate Security Precrisis: July 3, 1997-Aug. 14, 1998 0.960 0.942 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.918 0.936 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.941 0.954 Full sample: July 3, 1997-Oct. 29, 1999 0.938 0.946 Fannie Mae High-Yield Benchmark Corporate Period 0.976 0.527 Precrisis: July 3, 1997-Aug. 14, 1998 0.990 0.291 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.983 0.454 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.985 0.346 Full sample: July 3, 1997-Oct. 29, 1999 Period U.S. Treasury Precrisis: July 3, 1997-Aug. 14, 1998 0.987 Crisis: Aug. 14, 1998-Nov. 20, 1998 0.968 Postcrisis: Nov. 20, 1998-Oct. 29, 1999 0.961 Full sample: July 3, 1997-Oct. 29, 1999 0.970
Source: Author's calculations, based on data from Bloomberg, Goldman Sachs, and Merrill Lynch.
Notes: The table reports the correlations of weekly yield changes between the ten-year swap rate and the indicated index or security. Correlations with the Fannie Mae benchmark are limited to the period starting February 3, 1998. The swap rate is the semiannual fixed rate versus three-month LIBOR compiled by Bloomberg from various sources. The investment-grade corporate yield is the industrials ten-year A2/A yield from Bloomberg. The mortgage-backed security yield is a weighted-average, option-adjusted yield calculated by Goldman Sachs. The Fannie Mae benchmark yield is the on-the-run ten-year benchmark note yield from Merrill Lynch, via Bloomberg. The high-yield corporate yield is from Merrill Lynch's High-Yield Master Index, via Bloomberg.
The interest-rate swaps market is very active, with narrow bid-ask spreads. A market survey by the Federal Reserve Bank of New York (1998) found daily trading in U.S. dollar interest-rate swaps to be $22 billion per day in April 1998.(20) Turnover is thus considerably higher than it is in agency coupon securities, but less than it is in Treasury securities. Bid-ask spreads on active contracts reportedly are about 1 basis point, somewhat wider than those on active Treasury securities.
The liquidity of the swaps market is hindered by counterparty credit risk. Counterparty credit risk is the risk that one's counterparty in a swap defaults on its end of the agreement. The risk is an obstacle to liquidity because, by definition, it depends on the parties involved in a transaction. A dealer that has engaged in a swap contract and wants to unwind it either has to go back to the original counterparty, which may not want to unwind, or find a third party to take its side of the swap--one that is also acceptable to the original counterparty. To mitigate counterparty credit risk, some dealers execute swaps out of credit-enhanced subsidiaries and structure swaps so that they automatically unwind if a party's Aaa/AAA credit rating is lost.
The absence of an underlying fundamental asset is also an advantage of the swaps market. There is no supply limit on swap contracts and no need to borrow securities to go short, as an entity can enter into as many swap contracts as it wants. Specific issue concerns are also mitigated by the nature of swaps. The ability to create a swap combined with the fungible nature of the underlying cash flows precludes swaps with the same or nearly the same cash flows from trading at widely different rates.
Swaps are used as benchmarks for hedging positions taken in other markets, including the agency debt, corporate debt, and mortgage-backed securities markets. They are used as well for analytical and monitoring purposes in evaluating the performance of other fixed-income markets. Swap rates are also used as reference rates for forecasting, for example, the path of LIBOR.
Several features favor the interest-rate swaps market as a benchmark. As the underlying floating rate has credit risk, the performance of swaps is highly correlated with that of other spread products, and swaps have the potential to be a better hedge than Treasuries. The absence of an underlying asset allows for dealers to take unlimited long or short positions without having to worry about obtaining securities in the repo market. These same features mitigate security-specific issues that might cause a particular maturity swap to deviate sharply from the performance of the whole swaps curve.
However, counterparty credit risk is a feature that does not favor the swaps market as a benchmark. Such risk means that swaps created by different parties have different risks and are not perfectly fungible. Lack of fungibility adversely affects liquidity. Market participants have taken steps to mitigate the effects of counterparty credit risk, but it remains a hindrance to the market's liquidity and to the market assuming a larger benchmark role.
The country's improved fiscal situation raises questions about the U.S. Treasury market's benchmark status. If projected budget surpluses materialize, they could lead to a significant reduction in the Treasury market's size and to a deterioration in the market's liquidity and efficiency. A less liquid and less efficient market would represent a less useful benchmark of risk-free interest rates as well as a less useful benchmark for pricing and hedging positions in other markets.
The financial markets crisis of fall 1998 heightened investors' concerns about the Treasury market's benchmark role and provided insight into how the market may perform in the future. A flight to quality into Treasury securities caused yields between Treasuries and other fixed-income securities to diverge. A related flight to liquidity also led yields among similar Treasury securities to diverge. Market liquidity also declined, and the cost of borrowing securities through the repo market increased. After fall 1998, market conditions did not quickly return to precrisis levels, possibly reflecting a more general decline in fixed-income liquidity as well as a continued high demand among market participants for benchmark Treasuries.
Other fixed-income markets--including the agency debt, corporate debt, and interest-rate swaps markets--have demonstrated some of the characteristics that potentially make them suitable benchmarks for pricing and hedging purposes. Furthermore, the attributes that are favorable to a benchmark have been improving in the agency and corporate debt markets as benchmark debt issuance programs are expanding and steps are being taken to develop repo market activity. At this point, the agency debt and swaps markets are already assuming a limited benchmark role as hedging vehicles and as reference yields for market monitoring and analytical purposes.
(1.) For recent reviews of the U.S. Treasury market, see Dupont and Sack (1999) and Fabozzi and Fleming (forthcoming).
(2.) In contrast, floating-rate issues typically are priced relative to the London Interbank Offer Rate (LIBOR), the short-term rate charged among banks in the Eurodollar market. A recent issue of Daimler-Chrysler AG, for example, had a three-year floating-rate portion priced relative to three-month LIBOR along with five-year and ten-year fixed-rate portions priced relative to comparable Treasuries (Wall Street Journal 1999b).
(3.) Debt held by the public excludes $2.0 trillion held in U.S. government accounts. Debt figures are from the U.S. Congressional Budget Office (1999) and Treasury Bulletin (1999).
(4.) Significant debt management changes typically are announced at the Treasury's Quarterly Refunding Press Conferences. The press releases for such conferences are posted at http://www.treas.gov/press/releases. Also see U.S. General Accounting Office (1999) for a more extensive discussion of recent changes in Treasury debt management.
(5.) The buyback rules are described in detail in the Federal Register and are available at http://www.publicdebt.treas.gov/gsr/gsrbuyback.htm.
(6.) In fact, in February 2000, the Treasury announced a number of additional debt management changes at its Quarterly Refunding Press Conference, including a reduction in the issuance frequency of one-year bills from every four weeks to four times per year. This followed the release of a CBO budget and economic outlook in January 2000 that projected even larger surpluses over the next ten years.
(7.) See Bank for International Settlements (1999) for an analysis of the events of fall 1998.
(8.) The precrisis, crisis, and postcrisis time periods are defined somewhat arbitrarily. The precrisis period runs from July 1, 1997, through August 14, 1998--the Friday preceding the Russian effective default and ruble devaluation on August 17, 1998. The crisis period runs from the close of August 14, 1998, through November 20, 1998--the Friday after the Federal Reserve System's third and final fed funds target-rate cut of 1998, on November 17. The postcrisis period runs from the close of November 20, 1998, through October 29, 1999.
(9.) It is possible that such subperiod correlations mask a shift in the relationship among yield changes between periods. To test this possibility, we also estimated correlations between actual yield changes and the yield changes predicted for a security from a least-squares regression of that security's yield changes on Treasury yield changes for the preceding ten weeks. These correlations are similar to those reported in Table 1 and are therefore not reported separately.
(10.) Volatility was estimated on a daffy basis over the full sample period using a GARCH(1,1) model of on-the-run ten-year note yield changes. Predicted volatility from this model helps explain the variation in both bid-ask spreads and quoted depths. However, dummy variables representing the crisis and postcrisis periods remain highly significant explanatory variables, even after controlling for predicted volatility.
(11.) The comparable off-the-run yield is calculated as the yield predicted for the on-the-run security from a model of the yield curve estimated with off-the-run prices. The model is estimated using a flexible functional form proposed by Fisher, Nychka, and Zervos (1995) in which a set of simple functions (cubic splines) covering different maturity ranges are used to describe the zero curve. The model is estimated to fit Treasury bid prices, excluding the two most recently issued securities of a given maturity, securities with less than thirty-one days to maturity, callable bonds, flower bonds, and inflation-indexed securities.
(12.) The increased relative value of on-the-run securities also likely reflected the securities' increased specialness in the repo market. The relationship between Treasury security value and specialness is discussed and documented in Duffie (1996) and Jordan and Jordan (1997).
(13.) The predicted yields are estimated according to the process described in endnote 11. The median is estimated daffy for off-the-run notes and bonds with more than thirty days to maturity, excluding callable bonds, flower bonds, and inflation-indexed securities.
(14.) The premium afforded to liquid on-the-run securities may explain why some market participants started using off-the-run Treasury yields for pricing corporate securities and as market barometers (Wall Street Journal 1999a). Unfortunately, the same feature that may make off-the-run Treasuries a better gauge of Treasury market performance--their relative lack of liquidity--also makes them poor vehicles for hedging purposes as well as more susceptible to idiosyncratic price changes.
(15.) See Fabozzi and Fleming (forthcoming) for a recent review of the agency debt securities market.
(16.) Fannie Mae stated that "the liquidity of the benchmark notes combined with the outstanding credit quality should cause benchmark notes to be viewed by many investors as a higher yielding alternative to off-the-run Treasuries" (http://www.fanniemae.com/markets/debt/benchmark_prod.html). Freddie Mac indicated that "the fundamental characteristics of reference notes are designed to appeal to investors seeking alternatives to the declining supply of U.S. Treasury notes and bonds" (http://www.freddiemac.com/debt/html/borrowprog.html). Finally, the FHLBanks remarked that "TAP issues have many of the properties of U.S. Treasuries" (Federal Home Loan Banks 1999).
(17.) However, in January 2000, both the Chicago Board of Trade and the Chicago Mercantile Exchange announced plans to list agency note futures and options contracts.
(18.) In August 1999, a new issue of Private Export Funding Corp. was marketed in terms of Fannie Mae's benchmark ten-year note, reportedly the first private debt issue priced off an agency security (Wall Street Journal 1999c).
(19.) Freddie Mac, for example, announced a financing calendar in June 1999 (Freddie Mac 1999) and Fannie Mae announced a goal of $6-$8 billion issuance sizes for new benchmark notes in October 1999 (Fannie Mae 1999c).
(20.) Note that this is the average notional principal amount on which parties agreed to exchange interest payments, rather than the value of securities traded.
Bank for International Settlements. 1999. "A Review of Financial Market Events in Autumn 1998." October.
Bloomberg. 1999. "Ford Credit's $5 Billion Sale Taps Demand for Big Issues." October 21.
Bond Market Association. 1999. RESEARCH QUARTERLY (November).
Bureau of the Public Debt. 1999. "Summary of Public Debt Outstanding." October 31.
Duffie, Darrell. 1996. "Special Repo Rates." JOURNAL OF FINANCE 51, no. 2: 493-526.
Dupont, Dominique, and Brian Sack. 1999. "The Treasury Securities Market: Overview and Recent Developments." FEDERAL RESERVE BULLETIN 85, no. 12 (December): 785-806.
Fabozzi, Frank J., and Michael J. Fleming. Forthcoming. "U.S. Treasury and Agency Securities." In Frank J. Fabozzi, ed., THE HANDBOOK OF FIXED INCOME SECURITIES. 6th ed. New York: McGraw-Hill.
Fannie Mae. 1999a. "Using the Fannie Mae Bullet Benchmark Securities Yield Curve as a Market Pricing Reference." FUNDINGNOTES 4 (September).
--.1999b. "Fannie Mae Makes Enhancements to the Benchmark Securities Program." FUNDINGNOTES 4 (October).
--. 1999c. "Fannie Mae Announces Monthly Calendar for Issuance of Bullet Benchmark Securities Spanning Yield Curve." News release, October 21.
Federal Home Loan Banks. Office of Finance. 1999. "Investors and Dealers Embrace the New FHLBank Program." OF INTEREST (November).
Federal Reserve Bank of New York. 1998. "Foreign Exchange and Interest Rate Derivatives Market Survey: Turnover in the United States." September 29. Federal Reserve Bulletin. Various issues.
--. 1999. November.
Fisher, Mark, Douglas Nychka, and David Zervos. 1995. "Fitting the Term Structure of Interest Rates with Smoothing Splines." Board of Governors of the Federal Reserve System Finance and Economics Discussion Series no. 95-1, January.
Ford Motor Company. 1999. "Quarterly Report, 1999 2nd Quarter." News release, July 14.
Freddie Mac. 1999. "Freddie Mac Announces New Reference Note Financing Calendar." News release, June 8.
Jordan, Bradford D., and Susan D. Jordan. 1997. "Special Repo Rates: An Empirical Analysis." JOURNAL OF FINANCE 52, no. 5: 2051-72.
Treasury Bulletin. Various issues.
--. 1999. September.
U.S. Congressional Budget Office. 1999. "The Economic and Budget Outlook: An Update." July 1.
U.S. General Accounting Office. 1999. "Federal Debt: Debt Management in a Period of Budget Surplus." Report no. AIMD-99-270, September.
Wall Street Journal. 1999a. "Quirk in Yields Is Making Bonds More Attractive." February 2.
--. 1999b. "Treasurys' Trading Volume Is Down Sharply amid Budget Surplus, Weak Market for Bonds." August 17.
--. 1999c. "Bonds Sustain Rally on Low Inflation, Expectation of Fed Restraint on Rates." August 26.
When thinking about how different asset classes might take over benchmark status from Treasury securities, it is useful to look back at history. We have been here before. Twenty-five years ago, Treasury securities were the benchmark securities for the fixed-income markets at all maturities. Today, they are benchmarks only at the intermediate and long ends of the yield curve. In the late 1970s, the Eurodollar (LIBOR) cash market began to take over the benchmark status that Treasury bills had occupied. Starting in the mid-1980s, the Eurodollar futures market became the hedging and trading vehicle of choice for the entire short end of the market. While some of the factors that have led to the migration of the benchmark from Treasury bills to Eurodollars are not particularly relevant to the situation today, other lessons from that experience may be instructive for the issues we will face in the coming months.
One of the reasons--perhaps the reason why Treasury bills lost their benchmark status--is that they are not ideal hedging vehicles. Treasury bills are subject to very substantial supply shocks and the supply of bills is interest-inelastic. Thus, the Treasury market is a less efficient market than other fixed-income markets in which both supply and demand respond to changes in interest rates. Inelastic supply is also a feature of the markets for intermediate- and long-term Treasury coupon securities.
In the Treasury bill sector, inelastic supply has been exacerbated by large shifts in supply--month to month, year to year, and over the course of the business cycle. For example, if tax receipts are extraordinarily high, the supply of bills may fall dramatically in the spring, effectively decoupling the Treasury bill from private sector interest rates. As a result of uncertainties in supply, the bill market's benchmark status became vulnerable as soon as more liquid, private sector short-term securities became available.
A related issue, alluded to in Michael Fleming's paper, is the fact that risk-free assets like Treasury securities can be poor hedging vehicles for other fixed-income securities because they do not have the same credit risk characteristics. At the end of the yield curve, LIBOR and Eurodollar futures are inherently superior hedging vehicles (relative to Treasury bills) for most private securities because they incorporate a sort of generic private sector risk premium.
As an aside, I would like to note that what we commonly call the credit risk spread--the difference between private fixed-income yield and comparable Treasury yield--is actually only half of a credit risk premium. The other half is a supply effect reflecting the fact that supply in the Treasury market, particularly the supply of bills, is arbitrary and interest-inelastic, and thus creates a distortion. A lot of things that we historically have called a credit risk spread are in fact just the result of shifts in the supply of Treasuries.
While the market for Treasury coupon securities has some of the same attributes as the bill market, the coupon sector did not lose its benchmark status in the 1980s for two main reasons: market liquidity and the lack of deep alternative markets. First, the Treasury traditionally has worked to keep coupon sector supply as regular and predictable as possible. This (relatively) fixed supply schedule has been very important in developing liquidity in the Treasury market over the years. While no private sector borrowers would commit to a fixed supply schedule because it is not optimal for them, a fixed supply has paid off for the Treasury in terms of vastly increased liquidity and an accompanying liquidity premium. In addition, the fact that Treasury coupon supply is inelastic may have increased trading volume and liquidity over the years, because it leaves a small market inefficiency for traders in Treasury securities to arbitrage away. Thus, the Treasury may benefit from forgoing the opportunity to exploit inefficiencies in its own market.
Looking ahead, it is not clear that supply even in the coupon sector can stay predictable. That is precisely the reason why people are asking whether coupons, like bills, are going to cease to be the benchmark.
The other reason why Treasury coupon securities did not lose their benchmark status in the 1980s is that there were no deep alternative private markets. That has changed. With the improvement of information technology and credit monitoring techniques, a much broader array of private sector borrowers now issue fixed-income securities with long maturities.
I will now turn to why LIBOR and Eurodollar futures replaced bills as short-term benchmarks fifteen or twenty years ago. From there, I will draw lessons from what might happen if coupon Treasuries now begin to lose their benchmark status.
What were the factors that gave Eurodollars the advantage over other private sector alternatives? The first factor was pricing transparency, particularly the fact that the Eurodollar market had a published reference rate (LIBOR) that was commonly accepted. The British Bankers Association LIBOR fixings were crucial to the acceptance of LIBOR as a pricing standard--and, ultimately, when a futures contract was offered, as a transaction and hedging vehicle.
The second factor was that the Eurodollar futures contract was designed as a cash-settled rather than a deliverable contract. Previous attempts to develop short-maturity private sector alternatives--commercial paper futures, domestic CD futures--died because of the messiness of deliverability. For Eurodollars, the futures market capitalized on the success of the published reference rate--LIBOR--to create a hugely liquid vehicle that traded just as a derivative.
What does that tell us about the options for alternative benchmarks going forward? One general lesson is that the better match between the credit risk of Eurodollars and that of other short-term securities gave Eurodollars a leg up in liquidity (relative to Treasury bills), particularly in times of market stress. Also, the Treasury bill market did not disappear when Eurodollars took over the benchmark status. It kept thriving as a store of liquidity and as a place to put money in tough times. It continued to be an extremely cheap source of funding for the Treasury. In other words, the Treasury does not have to worry about losing the benchmark status. Treasury securities will still have a role to play as a safe haven; they will still have very low rates even if they cease to be a primary trading vehicle.
Going forward, swaps may have an edge as benchmarks over agencies in part because it would be very easy to develop a widely recognized and accepted swap reference series. There is already an International Swaps and Derivatives Association swap series that could be improved and promoted and could become the basis of organized futures trading. This has not happened yet, and attempts to make it happen thus far have not been successful. However, if an improved hedging vehicle that does not decouple from private sector instruments in times of stress is needed, it would not be hard to create such a reference rate.
Finally, I believe that the same lesson applies to agency securities. Agencies right now are a hot candidate to be benchmarks because their individual issue sizes come close enough to Treasury issue sizes to be liquid and tradable. However, history suggests that an agency futures contract based not on messy deliverability considerations but on an index of agency yields (with no balance-sheet implications for those who use it only as a hedging vehicle) is probably superior to a system in which individual cash agency securities are treated as the benchmark. The agencies would love to see this happen because such a system does not rely on large trading volume in the underlying security, but rather on the development of an index-based futures contract.
One of the major lessons learned from October 1998 concerned the use of Treasuries as hedging vehicles for private sector securities that have different credit risk as well as different supply characteristics. If we accept the notion that a derivatives instrument based on an index of securities is a desirable benchmark for hedging, then the corporate bond market is also attractive. If the corporate bond market develops an index made up of large issue sizes--such as Ford's global bonds--then you will have an ample supply of issuers trying to participate in that index. As a result, a futures contract on such an index could easily be traded, and might ultimately be the best hedging vehicle for those who underwrite corporate issuance.
Lou Crandall is the chief economist at Wrightson Associates.
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System.
This thought-provoking paper by Michael Fleming raises several interesting issues in light of my experience, and makes an effort to establish some empirical regularities relating to different benchmark securities. After a brief review of the paper's major conclusions, I will address a set of public policy issues that the paper raises.
MAJOR CONCLUSIONS OF THE PAPER
First, the premise of the Fleming paper is that the value of the Treasury market as a benchmark will be called into question by improved fiscal performance. This conclusion is itself predicated on a trend shift in productivity growth and greater fiscal restraint that will lead to extensive efforts to pay down debt over a protracted period.
Second, the paper contends that recent worldwide shocks and events including the Long-Term Capital Management (LTCM) crisis "heightened concerns about the Treasury market's benchmark role."
Third, the paper argues that increasingly there will be alternative benchmarks emerging for the pricing and hedging of securities, including the agency debt, corporate debt, and swaps markets. Much of this argument is based on the idea that these forms of debt are characterized by credit risks that will be more correlated with spread products and that these forms of debt will be a better hedge than Treasuries--despite disadvantages in such areas as market size and liquidity.
PUBLIC POLICY ISSUES RAISED BY THE PAPER
Although the Fleming paper presents some interesting empirical correlations, relationships, and trends, it leaves the reader asking several questions--all of which have a public policy implication and none of which are actually discussed that explicitly.
These questions include:
* What characteristics should a benchmark security actually have and, more basically, what do we mean by a "benchmark"?
* Is the premise of the paper, which suggests the need for new benchmarks versus a Treasury benchmark, actually relevant?
* Might it be that the Treasury market (on-the-run and off-the-run issues) actually functioned quite well during the fall 1998 crisis and during the run-up to Y2K in recent months?
* Are the recent changes relating to the repo market and the eligibility of agency debt as collateral in Federal Reserve System open market operations worth maintaining in light of the discussion of alternative benchmarks, or are there reasons why this would be dangerous public policy?
* What are some of the specific advantages and disadvantages of each form of alternative "market benchmark" noted in the paper?
* Can we expect systemic and other forms of risk to increase with the introduction or proliferation of many different benchmarks and with the advent of many types of trading formats--such as ECNs and the new E-bond market?
WHAT DO WE WANT IN A BENCHMARK SECURITY?
A benchmark is a concept that can have a variety of meanings. One definition used in portfolio management refers to a benchmark portfolio of securities against which performance can be measured. Another meaning refers to a benchmark security whereby the market determines what specific issue or form of security can serve in such a capacity. Several characteristics seem critical: the credit quality of the issuer must be very strong, the issue must be very liquid (transactions should not materially impact the price of the security), and the overall structure of the market for the contract or security in question must have what we might call "integrity." Therefore, the market for a benchmark security should have minimal prospects of being squeezed or cornered by participants.
Benchmark securities are also important for properly measuring and calculating the value of other securities in the same class or other financial contracts more broadly. Often, Treasury securities are useful because they reflect a riskless rate of return. As such, these securities can be compared with other nongovernment-backed securities subject to greater credit risks. In this way, Treasury securities help to define the shape of the credit curve by pinning down the overall level of the credit curve that all lenders and borrowers can see. It is important to note, however, that even Treasury spreads (and securities) reflect a large number of risks--including duration (or average life risks), financing risks, haircuts in repurchase-related transactions, and supply and demand pressures--for on-the-run and off-the-run issues.
Fleming's paper often tends to confuse the roles and functions of a benchmark security with hedging, pricing, and liquidity. Although these many aspects of a security or market can be interrelated, it is clear that markets are evolving in the United States to separate these risks. For example, the swaps markets are critical for hedging and immunizing against certain forex or interest rate risks. However, swap rates themselves are based on underlying cash flows on fixed-income instruments or foreign exchange contracts--in spot or cash markets. Moreover, credit counterparty risk in swaps--thanks to International Swaps and Derivatives Association (ISDA) conventions--is being reduced as a form of variation, and initial margining is beginning to make these contracts similar to exchange-traded contracts. Hence, swaps or other derivatives are somewhat difficult to think of as benchmark securities under the kind of definition one might normally use.
The above considerations highlight the concern that the paper needs to be a bit more precise in defining what is meant by a benchmark security. It seems clear that a benchmark security should above all be liquid. Such a security should have simple properties and should be capable of being used as a building block in valuing other, more complex financial contracts or securities. In this context, it is critical that this financial contract and the market in which it trades have integrity, as I indicated above.
WHY WILL THE TREASURY BENCHMARK CEASE TO EXIST?
One premise of the Fleming paper rests on the assumption that a business cycle as we know it will not be present in the next decade. Instead, economic growth combined with a small but persistent trend shift in productivity to about 1.7 to 2.0 percent per annum will generate very large fiscal surpluses. If the trend shift were to be larger (all else being equal)--as implied by some recent studies--then the speed at which the size of the Treasury market would be reduced would accelerate. Such assumptions have always proved questionable, as explicitly mentioned in Office of Management and Budget and Congressional Budget Office projections. Changes in tax and expenditure policies as well as possible modifications to the U.S. health and pension systems could greatly alter many such forecasts. In addition, despite the unprecedented strength of the current business cycle, a slowing in economic growth needs to occur--given current rates, which are close to 5 percent in real terms--with implications for future surpluses.
Even if one feels that the U.S. debt-to-GDP ratio and the absolute debt level will fall dramatically, there are many actions that could be taken to preserve Treasury securities as a liquid benchmark. Among these would be a number of simple steps that, if combined, could act as a powerful force to improve the depth and liquidity of the Treasury market.
These actions could include:
* Further efforts could be made to reduce the effective Federal Reserve holdings of on-the-run Treasury securities. Here one can ask if the current holdings are justified from the vantage point of monetary control versus the obvious fiscal gains associated with holding a greater proportion of off-the-run Treasury securities, given the Federal Reserve's role as fiscal agent of the Treasury.
* The selective reopening of key Treasury issues or the removal of issues from the calendar and the concentration of issues to create liquid benchmarks, which has already begun, could be continued or intensified. The Canadian authorities and many other treasuries throughout the world are adopting this type of strategy.
* A reevaluation of the issuance of Treasury Inflation Protection securities could be conducted. There is a variety of other, more liquid contracts trading that could be used to gauge inflation expectations.
* The investment guidelines for the Social Security trust fund could be changed to permit a somewhat greater range of investments, which would free up room for private market participants to gain greater access to the on-the-run and off-the-run Treasury markets. Ginnie Mae mortgage-backed securities, Fannie Mae mortgage-backed securities, and Federal Home Loan Mortgage Corporation mortgage-backed securities (subject to proper structures) are examples. The investment guidelines would have to be specified very carefully and stress capital preservation. Such activities versus equity investment would certainly not seem unsound--particularly in the case of Ginnie Mae securities.
* In the extreme case--where the supply of Treasury securities becomes very small and where the Federal Reserve feels uncomfortable undertaking repo transactions based on the use of agency or other debt as collateral and sees value in a Treasury market--other alternatives could be contemplated. Specifically, the Federal Reserve could act to issue debt that it backs and simultaneously sterilize this debt issuance by originating an asset. Under these circumstances, changes in the U.S. legislative framework would be needed, as the central bank presently can act only as the fiscal agent of the Treasury. This idea presumes that having a government Treasury benchmark security is important enough to change the nature of the relationship between the fiscal authorities and the central bank. Such arrangements are not at all uncommon in both developed and developing countries throughout the world. This alternative is obviously not an option that needs to be considered in the short term.
The above considerations highlight the notion that there need not be a rapid deterioration in the effectiveness of the U.S. Treasury market as a benchmark for either on-the-run or off-the-run Treasury securities.
THE FALL LTCM CRISIS AND THE TREASURY MARKET
Fleming's paper does a good job of documenting the complex issues raised by the crisis in 1998 and the problems of Long-Term Capital Management, as well as the total seizing up of credit markets and the flight to quality into on-the-run Treasuries. However, it is very difficult to see how those events call into question the effectiveness of the Treasury market as a benchmark.
First, even prior to the crisis, spreads between swaps and off-the-run Treasuries were wide.
Second, and more importantly, the widening of yield spreads between on-the-run and off-the-run Treasuries is in fact the kind of reaction one can expect in a generalized market panic, where many counterparties were unclear as to the extent of risks being undertaken.
Third, recent movements in swap and other spreads have had more to do with large anticipated borrowing requirements prior to Y2K and less to do with systemic risks.
Perhaps most importantly, the LTCM crisis illustrates the fact that the Treasury market enabled markets to absorb an unprecedented shock. The lessons, in my view, have much more to do with the risk management techniques being used and the inability of models and techniques such as value-at-risk to account properly for extreme cases of liquidity risk, than they have to do with defects in the Treasury market per se.
Finally, the role of hedge funds and prop desks in providing liquidity to the Treasury market is also important. Ironically, this will require very careful changes in disclosure policies, as the very nature of trading in any market requires that the participants have no knowledge of the size of the other participants' positions. Moreover, recommendations relating to the disclosure of positions to regulatory agencies could also be problematic depending on how and for what purpose such information is used. It is very clear that the credit evaluation process used in lending to hedge funds like LTCM is among the more critical areas where improvements have been and will continue to be made.
EXTENDING Y2K-RELATED CHANGES AND BENCHMARKS
As part of the effort to mitigate problems related to Y2K monetary authorities, the United States undertook a number of actions, including a broadening of the set of securities that can serve as eligible collateral in repos with the Federal Reserve. It is worth noting that these changes in procedure will be reviewed to see if they should be kept in force beyond April 2000.
Although not discussed in Fleming's paper, the implications of allowing most forms of agency debt to be eligible collateral in repos with the Federal Reserve represent a significant step. This action provides added liquidity and credibility to these markets and might be viewed by market participants as enhancing the liquidity of the special benchmark security programs initiated by the agencies.
Ironically, and in contrast with the argument above, in many emerging markets questions would typically be raised if the monetary authorities were thought to be taking on credit risk by dealing in these securities. In the U.S. context, some would argue that this is a kind of back-door method for these agencies to assert that their securities are in fact backed by the central bank and U.S. government, thereby lowering funding costs. Such arguments might apply even if the U.S. authorities made haircuts when such paper is pledged as collateral. The public policy issues surrounding extension of this policy would be worthy of study, either separately or in Fleming's paper. One could even look at the impact on the liquidity of the agency and other markets that these policies have implied to date.
ALTERNATIVES TO THE TREASURY BENCHMARK
Fleming suggests that agency debt, swaps, and corporate debt markets will all become more important as benchmarks. Evidence does suggest that these markets are growing quickly, and agencies have been quick to see that their funding costs can be reduced through careful and strategic placements of debt, including the use of benchmark notes (for example, Fannie Mae) or reference notes (Freddie Mac). In the paper, some of the arguments made for the effectiveness of these benchmarks rest on their correlation with the U.S. Treasury market. In this context, much of the data in the paper are a bit confusing because at times it is unclear if the correlation coefficients are derived on the basis of first differences or levels when the paper refers to the correlation of daily yield changes. In other sections on market liquidity, it is unclear that proper account has been taken of seasonal impacts. In sum, I have some trouble seeing how the empirical work done in the paper supports the contentions made about the effectiveness of specific benchmarks.
The fact is that agency debt carries credit risk, and its correlation with spread products does not automatically make such debt a better hedge, as is claimed in the paper. Rather, the issue here is which financial contracts provide the best means at the lowest costs, including liquidity and other risks of hedging specific forms of risk. In this context, the swaps market offers advantages under many circumstances if such contracts are ISDA-conforming relative to agency debt.
Finally, the adequacy of each of these markets must also be assessed in terms of the credit quality of the underlying issuer and the implications for market integrity and systemic risks. Here, even the agency benchmark market could be viewed with some question. For example, the agencies have to increase the size of their mortgage loans and make other changes in their asset-side origination policies to be able to meet continually their supply commitments on benchmark issues. In addition, as interest rates continue to rise and as mortgage loan origination and refinancing drop off, credit quality could in effect be hurt and the integrity of the new benchmarks could be damaged.
In sum, many of the new benchmark securities may be subject to credit quality issues that are business-cycle-dependent.
THE INTERNET, E-BONDS, AND BENCHMARK SECURITIES
A last area not addressed by the Fleming paper, but a fruitful area for future research, is the confluence of risks that may start to be created by internet banking and the much more active use of electronic trading formats (for example, Trade Web). These risks would apply to the market for new bond issues as well as to secondary-market and after-market trading coupled with the development of many forms of portfolio benchmarks and many different benchmark securities.
These developments will present great challenges in the design of regulations for the Securities and Exchange Commission and even for the Federal Reserve. Although such technological developments can create tremendous scope for reductions in transaction costs and can reduce the operational costs faced by financial institutions on the sale side of the business, the maintenance of market integrity could become challenging.
It would not take much imagination to envision situations in which a shock leads to a flight to quality and many benchmark securities begin to fall in price simultaneously. Importantly, the transmission of a shock to asset-price movements and the extent of volatility might be much more rapid as technological advances in trading formats become more commonplace. In such cases, the authorities' latitude in ways to deal with the problem might be limited--purely because the speed of reaction necessary would not be feasible. More generally, issues relating to operational and systemic risk would become important.
Thomas C. Glaessner, formerly of Soros Management Fund LLC, is now the lead financial economist at the World Bank.
The author thanks Alan Boyce and Duncan Hennes for helpful discussions in the preparation of these comments. The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System.
Much of what was discussed in Michael Fleming's paper and those that preceded it was, in my opinion, interesting, but methodologically flawed. Of course, the data in the papers are all correct, but I would like to present an alternative view that explains what happened to the fixed-income markets in the fall of 1998 as well as shows that the concern over Treasuries' benchmark status is sort of a black flag without much meaning.
To begin, there is not enough historical perspective in these papers. We have been here many times before. The major problem we encounter is that Treasuries are a poor hedging vehicle. One reason for this problem is that people always assume that the representative investor is long securities and wants to short on-the-run Treasuries as a hedge. In the first half of 1998, the spread between on-the-run and off-the-run Treasuries was almost zero. The question, then, is who would want to be long an off-the-run security and short an on-the-run security at a yield spread approaching zero? This is a position in which a trader will make no money if things go well (if spreads remain narrow) and one in which a trader will get hurt badly if spreads widen. Moreover, history tells us that under such conditions there is a very big possibility that a large "event" will cause spreads to widen. This is exactly what happened in the second half of 1998.
Part of the problem leading up to fall 1998 was the poor use of econometrics, particularly by certain hedge funds. Modern risk management systems rely heavily on calculating value at risk and other measures of potential losses using statistics based on data from the recent past. Of course, these statistics cannot evaluate gains and losses for events that did not happen. As a result, if we develop a value-at-risk statistic during relatively stable times with narrow spreads, many spread trades will look relatively safe, and market participants--in this case, certain hedge funds--will start investing in them on a heavily leveraged basis.
Furthermore, the increased speed of trading and data analysis in recent years has made this problem more complex. All traders use essentially the same methodology to evaluate risk. In addition, everyone analyzes the same data on a daily basis. Thus, everyone conducts the same basic trades and arbitrages. In such a marketplace, when a large (negative) shock to the system occurs, the risk management systems indicate that traders should liquidate their positions at approximately the same time. By doing so, of course, the traders push prices down even further, which causes them to liquidate even more positions, and so on.
This situation was complicated last year by the structure of the Treasury repo market. First, this market is, at least during normal times, almost 100 percent leveraged. This is a poorly understood fact of the market. Dealers themselves do not pay any margins, and market making is so competitive (in good times, at least) that anyone making large trades can shop around until a dealer is found who is willing to finance at nearly 100 percent. During the 1998 crisis, some of the leverage disappeared. Large traders, such as hedge funds and relative-value firms, very quickly were asked to put up 2 percent margin, rather than almost zero. As a result, many relative-value trades, which had looked attractive when financed at 100 percent, became de minimis trades when financed at 98 percent. This issue highlights the nature of arbitrage: trading huge amounts of securities for a miniscule spread on a highly leveraged basis. A small change in the cost of leverage will force traders out of arbitrage because of their risk management constraints.
By the way, one reason why margins rose and leverage fell was that the dealer community and the bank community had exactly the same kinds of trades as the hedge funds did. When dealers and banks began to post their own spread losses, their risk management systems indicated that they should reduce their positions and their lending, which raised margins.
My conclusion is that there was no flight to quality into Treasuries in the fall of 1998: instead, there was a liquidation of short Treasury positions by massively leveraged hedge funds. These actions drove spreads up to such an extent that other market participants, many of whom had entirely different trading strategies, were forced to sell or close positions when their value-at-risk models indicated that their hedges had deteriorated. The irony is that the existence of a Treasury benchmark worsened the situation. As Treasury yields were pushed lower, all spreads widened, making even more positions unprofitable and causing dealers to raise margins on repos, which in turn caused even more liquidation of short Treasury positions, pushing Treasury yields even lower, and so on.
As a result of these events, it is important to think about hedging within a more generic framework. The basic risk borne by the financial marketplace--that is, the dealer community--is not "level" risk, but correlation risk. For example, how does the yield spread of one country move against that of another country? What happens to the shape of the entire yield curve under different scenarios? How does one price a Bermuda swaption in the United States? For these risks, the key criterion is the correlation between asset-price movements and spreads. Therefore, Treasuries often end up being the worst hedges for such complicated risks, in part because large shocks can significantly change their correlations with other asset prices.
I believe that the usefulness of employing Treasuries for hedging purposes has already passed. I have to agree with those who argue that swaps, as the market is now structured, are almost risk free and in some ways probably less risky than Treasuries. Certainly, the credit risk in the underlying LIBOR is very small, because poorly performing banks are dropped out of that index by the British Bankers Association. In addition, nearly all swap transactions are now (or soon will be) marked to market daily. Thus, no matter what the underlying credit problem is, a trader will have at most one day's price movement risk on a swap, which is essentially the same risk that traders have on Treasury and repo transactions. That is to say, when you buy a Treasury issue or you do a repo, your credit risk is the risk that the dealer might not be around the next day to deliver the security. Furthermore, in good times at least, the Treasury repo market--like the swaps market--is almost 100 percent leveraged.
It seems clear to me that a benchmark futures contract based on LIBOR swaps will be able to replace Treasuries. In this sense, the United States will be following Europe. When the European swaps market first began to develop, I recall visiting European institutions and explaining how we priced instruments from government benchmarks. People there found this practice surprising--nobody knew what a government benchmark was. The institutions traded their securities from swaps and futures benchmarks rather than from governments. Today, the European marketplace has the largest futures trading in the Eurex, far surpassing U.S. futures contracts and fixed-income securities. Furthermore, because the swaps market in the euro is the universal market, swaps spreads usually trade below government spreads. This is completely rational for the euro because swaps are far more liquid than instruments such as European government bonds.
Interestingly enough, the benchmark shift has already started in the United States. The thirty-year Treasury bond is no longer the lead contract for the U.S. futures markets. About four weeks ago, the ten-year note futures became the dominant futures contract in the United States, trading more open interest and volume than the thirty-year bond futures did. In addition, corporate bond debt and mortgage-backed debt traders are now hedging their collateral with interest-rate swaps. This is certainly not a risk-free game and, looking forward, we are likely to see the swaps market change its characteristics. For example, swap rates may be influenced by mortgage prepayment risk, if mortgage-backed securities are hedged in swaps first and then filtered through to the Treasury market.
In point of fact, the benchmark status of the Treasury market has been changing over the past ten to twelve years. Within five to ten years, it seems almost certain that we will have a swaps-based financial marketplace, where only the cash flows will matter, and where market participants will not be concerned with how the flows are bundled. This scenario will be an improvement over the uncertain supply conditions that often drive Treasuries and, more importantly, the Treasury repo market.
In fact, the Treasury repo market is unique: no other country has or is developing the kind of liquid repo markets that we have in the United States, and these markets certainly do not serve as benchmarks. Elsewhere, when a government security position is financed, it is done at about the swaps or the LIBOR rate. Yet it is precisely this phenomenal institutional repo system in the United States that drives the market for Treasuries and, as such, sometimes makes Treasuries appear to be so strange and distorted in their relationships with other instruments. If the Treasury market's benchmark status is changed, and there is much less need to borrow and lend Treasuries directly, I believe we will see a much more stable environment for trading and hedging. In addition, the combination of Treasuries being "risk-free" and being a benchmark is detrimental to hedging, because only the government can actually borrow risk-free. By changing benchmarks, we will alleviate some of that hedging problem.
Stan Jonas is a managing director at FIMAT Futures USA, Inc.
The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System.
Michael J. Fleming is a senior economist at the Federal Reserve Bank of New York.
The author thanks Peter Antunovich, Robert Elsasser, Kenneth Garbade, Charles Jones, Frank Keane, Jim Mahoney, Frank Packer, Adam Posen, Tony Rodrigues, and Federal Reserve Bank of New York seminar participants for helpful comments. The research assistance of Daniel Burdick is gratefully acknowledged. The views expressed are those of the author and do not necessarily reflect the position of the Federal Reserve Bank of New York or the Federal Reserve System.
Please note: Some tables or figures were omitted from this article.