Secretary
Securities & Exchange Commission
450 5^{th} Street, N. W.
Stop 109
Washington, D.C. 20549
May 4, 1998
RE: Comment Letter to Net Capital Rule  Fixed Income Proposed Rule
Release No. 3439455
File No. S73197
Dear Jonathan Katz:
The Securities and Exchange Commission has proposed amendments to 17 CFR 240 Rule 15c31 in Release No. 3439455. If the proposed amendments are adopted, the capital charges or "haircuts" associated with certain types of interest rate instruments would be changed. These instruments include government securities, investment grade nonconvertible debt securities, certain mortgagebacked securities, money market instruments, and debtrelated derivative instruments.
We would like to comment on the part of the fixed income proposed rule concerning the "deep discount bond." ^{1} Our analysis focuses on the hedged portfolio risk. Our results are shown below in Tables 1.
For illustrative purposes, Table 1 shows several possible hedged portfolios that are constructed from combinations of 30year coupon TreasuryBond and 15, 20 and 30year zero coupon Treasury securities. These hedged portfolios are grouped into zeros hedged with Tbonds and zeros hedged with other zeros. All long and short positions are for equal market values.
The Commission has traditionally chosen haircuts to ensure that the broker dealer will be able to realize the value after haircuts 95% of the time if it takes one month to unwind the portfolio. The current haircuts approximately equal two standard deviations of the portfolio’s monthly return. We would like to propose an alternative standard for deep discount Treasury securities. We believe these securities are highly liquid and even a large portfolio can be liquidated in one to two trading days. Therefore, we propose the oneweek portfolio return at the 95% confidence level as the appropriate standard for measuring risk from unwinding a portfolio for these highly liquid securities. Using a oneweek risk standard will generally have the effect of lowering the proposed haircuts. Table 1 shows the upper bound of the 95% confidence interval of onemonth and oneweek portfolio returns. Our discussion below is based on the oneweek portfolio risk.
In addition, the proposed haircuts do not reflect the relative risks of the hedged portfolios. For example we can select two hedged portfolios that have similar proposed haircuts, yet have very different probable losses. A hedged portfolio consisting of a long 15year zero and a short 20year zero coupon Treasury security (example 4 in Table 1), and a hedged portfolio consisting of a long 20year zero and a short 30year zero coupon Treasury security (example 6), have similar proposed haircuts of 7.8% and 7.5%, respectively. However, these two portfolios have very different probable losses, 2.68% and 5.54%, respectively, based on a oneweek standard. As a second example, we can select two hedged portfolios that have similar probable losses, yet have very different proposed haircuts. A hedged portfolio consisting of a short 20year zero and a long 15year zero coupon Treasury (example 4), and a hedged portfolio consisting of a short 30year Treasury bond and a long 15year zero coupon Treasury (example 1), have similar probable losses of 2.68% and 2.46%, respectively. However, these portfolios have very different proposed haircuts of 7.8% and 0.3%, respectively.
This phenomena may arise because of the manner in which the deep discount securities have been assigned to zones and subzones in the proposed rule. For instance, the proposed haircuts for hedged portfolios containing deep discount securities from different zones or subzones generally are greater than securities from the same zones or subzones, but the probably losses are not always greater. The proposed haircuts for hedged portfolios containing deep discount securities from the same zone or subzone generally are lower than the proposed haircuts for positions containing securities from different zones or subzones; however, the probable losses are not always lower.
We recommend that the Commission reexamine its haircuts concerning deep discount Treasury securities. We provide this analysis to the Commission which we believe demonstrates that the portfolio risk of deep discount bonds are not always consistent with proposed haircuts. However, our sampled data is limited to a ten year period and to zero coupon bonds. We recommend further empirical analysis incorporating longer history of price data and a fuller universe of deep discount securities prior to finalizing the Commission’s haircuts on deep discount bonds.
Table 1 Hedged Portfolio Returns* (9/87 to 3/98)  
Example Portfolios 
SubZones 
Portfolio Description 
Proposed Haircut 
Monthly Gain/Loss at 95% Confidence Level (%) 
Weekly Gain/Loss at 95% Confidence Level (%) 
TBond Hedged Portfolios:  
Example 1 
Zone 4 (xiii) v. 
30Year TBond Short v. 
0.3% 
5.10% 
2.46% 
Zone 4 (xiii) 
15Year Zero Long 




Example 2 
Zone 4(xiii) v. 
30Year TBond Short v. 
7.8% 
6.72% 
3.25% 
Zone 5 (xiv) 
20Year Zero Long 




Example 3 
Zone 4(xiii) v. 
30Year TBond Short v. 
10.8% 
14.44% 
6.98% 
Zone 5 (xv) 
30Year Zero Long 




Zeros Hedged with Zeros Portfolios:  
Example 4 
Zone 4 (xiii) v. 
15Year Zero Long v. 
7.8% 
5.55% 
2.68% 
Zone 5 (xiv) 
20Year Zero Short 




Example 5 
Zone 5 (xv) v. 
30Year Zero Long v. 
10.8% 
13.68% 
6.61% 
Zone 4 (xiii) 
15Year Zero Short 




Example 6 
Zone 5 (xv) v. 
30Year Zero Short v. 
7.5% 
11.14% 
5.54% 
Zone 5 (xiv) 
20Year Zero Long 




* 
See supporting information on attached analysis file: Zeros.xls which is available upon request.  
Sources: 
Data are from Bloomberg for 9/11/1987 to 3/31/1998 period. Coupon Treasury security yields are based on Constant Maturity Treasury yields. Zero coupon Treasury yields are based on the coupon portion of the STRIP.  
Notes: 
Portfolio returns are shown for the upper bound of the 95% confidence interval. See attached file Zeros.xls for the lower bound. Monthly and weekly gains/losses at 95% confidence level are 1.96 standard deviations of monthly and weekly returns. Standard deviations of monthly returns are calculated by multiplying standard deviations of daily returns by the squareroot of 21.4 (4 weeks’ of trading days). Standard deviations of weekly returns are calculated by multiplying daily returns by the squareroot of 5 (1 week of trading days). Daily returns are calculated by multiplying daily yield changes by the median modified duration for the sampled period. 
If you have any questions regarding this letter, please do not hesitate to call me at (212) 8725671. Jennifer Yoon of our Economics Consulting Services group provided assistance in this analysis, and you may also feel free to call her directly at (202) 9742036.
Very truly yours,
KPMG Peat Marwick LLP
Edward H. Jones, Partner
Attachment: File name: zeros.zip; original file name: Zeros.xls (ZIPped Microsoft Excel95 file; use PKZip to decompress)
(This additional supporting analysis is available upon request)
[1] The definition of a deep discount bond is given in footnote 29 of the proposed amended rule as follows: "Deep discount bonds are defined generally as Fixed Income Products that either do not pay interest or are priced at 50% or less of their par value." (See paragraph (c)(2)(vi)(A)(4)(iv) of the Proposed Amendments.) In our analysis, we have used zero coupon bonds as a proxy for all deep discount bonds.