Debt Management: Treasury Has Refined Its Use of Cash Management Bills but Should Explore Options That May Reduce Cost Further

Debt Management: Treasury Has Refined Its Use of Cash Management
Bills but Should Explore Options That May Reduce Cost Further
(30-MAR-06, GAO-06-269).

One result of persistent fiscal imbalance is growing debt and net
interest costs. Net interest is currently the fastest-growing
"program" in the budget and, if unchecked, threatens to crowd out
spending for other national priorities. This report was done
under the Comptroller General's authority. GAO examined the
Department of the Treasury's (Treasury) growing use of
unscheduled short-term cash management bills (CM bills).
Specifically GAO (1) describes when Treasury uses CM bills and
why, (2) describes the advantages and disadvantages of CM bills,
(3) describes steps taken by Treasury to reduce the overall
borrowing costs associated with CM bills, and (4) identifies
possible options Treasury could consider to reduce the use and
cost of CM bills further.
-------------------------Indexing Terms-------------------------
REPORTNUM: GAO-06-269
ACCNO: A50497
TITLE: Debt Management: Treasury Has Refined Its Use of Cash
Management Bills but Should Explore Options That May Reduce Cost
Further
DATE: 03/30/2006
SUBJECT: Borrowing authority
Cash basis accounting
Cash management
Debt subject to statutory limitation
Financial analysis
Interest
Lending institutions
Fiscal policies
Policy evaluation

******************************************************************
** This file contains an ASCII representation of the text of a **
** GAO Product. **
** **
** No attempt has been made to display graphic images, although **
** figure captions are reproduced. Tables are included, but **
** may not resemble those in the printed version. **
** **
** Please see the PDF (Portable Document Format) file, when **
** available, for a complete electronic file of the printed **
** document's contents. **
** **
******************************************************************
GAO-06-269

* Report to the Secretary of the Treasury
* March 2006
* DEBT MANAGEMENT
* Treasury Has Refined Its Use of Cash Management Bills but Should
Explore Options That May Reduce Cost Further
* Contents
* Results in Brief
* Background
* Objectives, Scope, and Methodology
* Treasury Uses CM Bills to Help Fill Cash Financing Gaps
* Large Cash Outflows Occur at the Beginning of the Month
* Large Receipts Often Occur in the Middle of the Month
* Regular Bill Issuances Alone Are Not Sufficient to Fill Cash
Financing Gap
* Treasury Has Increasingly Used CM Bills to Fill Frequent
Cash Financing Gaps
* Treasury Also Uses CM Bills to Help Smooth Cash Flows during
the April Tax Season
* Treasury Does Not Control the Timing of All Cash Flows
* Treasury Also Relies on CM Bills When Approaching the Debt
Ceiling
* CM Bills Cannot Be Used in All Circumstances and Other
Alternatives No Longer Exist or Are Limited
* CM Bills Generally Had Higher Yields than Outstanding Bills of
Similar Maturity but Lower Borrowing Costs than Currently
Available Alternatives
* CM Bill Yields Closely Track Other Short-Term Rates
* CM Bill Yields Were Generally Higher than Yields on
Outstanding Bills of Similar Maturity
* CM Bill Yields Were Similar to Yields on Overnight
Repurchase Agreements
* Our Analysis Identified Some Features That Affect the Yield
Differential
* The Amount of CM Bills Issued Was Positively Correlated
with the Yield Differential
* Issuing CM Bills in Multiple Tranches May Increase the
Yield Differential but Reduce Overall Borrowing Costs
* CM Bills Maturing on Large Tax Payment Dates May Have
Lower Yield Differentials
* Yield Differential Rises and Falls with Short-term
Yields
* CM Bills Are Innately Less Predictable than Regular
Bills
* Most Measures We Examined Suggest CM Bill Auctions Perform
Less Well Than 4-week Bill Auctions
* CM Bills Are Generally Less Costly than Currently Available
Alternatives
* CM Bills Are Generally Less Costly than Running Higher
Cash Balances
* CM Bills Are Generally Less Costly than Issuing
Longer-term Regular Securities
* Treasury Has Refined Its Use of CM Bills but Costs Have Begun to
Increase
* Treasury Initially Reduced the Amount of CM Bill Issuance
Following the Introduction of the 4-week Bill
* Issuing CM Bills in Multiple Tranches Has Helped Reduce
Borrowing Costs
* Borrowing Costs Declined Dramatically in 2002 but Have Since
Increased
* Options Exist That May Reduce the Use and Cost of CM Bills
* Better Aligning Cash Flows Might Reduce Cash Financing Gap
* Increasing Earnings on Cash Balances Would Reduce the Costs
Associated with Running Higher Cash Balances to Avoid
Issuing CM Bills
* Shorter-term Regular Bills Might Reduce the Need for Some CM
Bills
* Enhancing Transparency of CM Bill Auctions May Reduce CM
Bill Yields
* Relaxing the 35-percent Cap on CM Bill Auction Awards May
Reduce CM Bill Yields
* Exploring Other Countries' Short-term Cash and Debt
Management Practices May Provide Useful Insights
* Conclusion
* Recommendation for Executive Action
* Agency Comments
* Detailed Methodology and Findings of Statistical Analysis of Cash
Management Bill Yield Differential
* Previous Research Based on Earlier Data Suggests Some CM Bill
Features May Reduce Costs
* Comparing CM Bill Yields with Secondary Market Yields on
Similar-maturing Treasury Bills
* The Introduction of the 4-week Treasury Bill Reduced CM Bill
Maturities and Their Relative Importance in Short-term Financing
* Our Study Focused on CM Bills Issued during Fiscal Years
2002-2005
* Yield Differential Is Positively Correlated with Level of
Treasury Bill Yields
* Yield Differential Is Positively Correlated with Auction Amounts
* Maturity on or around Large Tax Payment Dates Reduces Yield
Differential
* CM Bills Issued in Multiple Tranches May Require Higher Yield
Differentials
* Yield Differential Is Not Affected by Timing of Issuance
* Several Findings from Earlier Studies Are No Longer Supported
* Selected CM Bill Features Reduce Yields but Opportunities for
Additional Exploitation of These Features May Be Limited
* Analysis of CM Bill Auctions
* Trading Activity Was Lower Prior to CM Bill Auctions than 4-week
Bill Auctions
* Yield Treasury Paid on CM Bills Generally Reflected Market
Information at the Time of the Auction
* CM Bill Auctions Are More Concentrated and Have Fewer Awarded
Bidders than 4-week Bill Auctions
* More Common Performance Measures Indicate High Demand for CM
Bills but Greater Uncertainty in CM Bill Auctions Compared with
4-week Bill Auctions
* Selected Bibliography
* Comments from the Department of the Treasury
* GAO Contacts and Staff Acknowledgments

Report to the Secretary of the Treasury

March 2006

DEBT MANAGEMENT

Treasury Has Refined Its Use of Cash Management Bills but Should Explore
Options That May Reduce Cost Further

Contents

Tables

Figures

March 30, 2006Letter

The Honorable John W. Snow Secretary of the Treasury

Dear Mr. Secretary:

One result of persistent fiscal imbalance is growing debt and net interest
costs. Net interest on the federal debt is currently the fastest-growing
"program" in the budget. The Congressional Budget Office (CBO) projects
that interest cost will increase by more than 50 percent over the next 5
years. If unchecked, interest spending threatens to crowd out spending for
other national priorities such as educating our children, safeguarding the
environment, and ensuring national security.1

The Department of the Treasury's (Treasury) primary debt management goal
is to finance the government's borrowing needs at the lowest cost over
time. To meet this objective, Treasury issues debt through auctions in a
"regular and predictable" pattern across a wide range of securities
including nominal bills, notes, and bonds, and Treasury
Inflation-Protected Securities (TIPS).2 According to Treasury, because
investors and dealers rely upon the routine availability of Treasury
securities they tend to pay a slight premium, which lowers Treasury's
borrowing costs.

Since 1975, Treasury has supplemented its regular and predictable auctions
with unscheduled cash management bill (CM bill) auctions to finance
short-term cash needs while providing short notice to market participants.
However, Treasury officials and researchers indicated that Treasury pays a
premium on CM bills in part because of their irregularity and short-term
nature. We examined Treasury's use of CM bills to provide insights into
options Treasury might consider to reduce CM bill use, the premium paid on
CM bills, and ultimately government borrowing costs.

This report is part of GAO's ongoing work on Treasury's cash and debt
management practices and was done under the Comptroller General's
authority. We examined general patterns of CM bill issuance over the last
10 fiscal years-fiscal years 1996-2005, but focused on the last 4 fiscal
years after Treasury introduced a new regular 4-week bill in 2001 to
reduce the use of CM bills. The objectives of this report are to (1)
describe when Treasury uses CM bills and why, (2) describe the advantages
and disadvantages of CM bills, (3) describe steps taken by Treasury to
reduce the overall borrowing costs associated with CM bills, and (4)
identify possible options Treasury could consider to reduce the use and
cost of CM bills further.

Results in Brief

Since 2002, Treasury has increasingly issued CM bills to fill intramonth
cash financing gaps that regular bills alone cannot fill. Treasury issues
CM bills in the beginning of the month to raise cash for Social Security,
federal retirement, and other scheduled payments. Treasury typically sets
the maturity of CM bills to the 15th of the month when notes are issued
and tax payments are received. Treasury has also relied on CM bills when
it is nearing the debt ceiling to help it pay the government's bills while
keeping debt under the statutory limit.3

CM bills provide Treasury flexibility to obtain cash outside its regular
borrowing schedule, but Treasury pays a premium for doing so. During our
study period (fiscal years 1996 through 2005) Treasury paid a higher yield
on most CM bills than outstanding bills of similar maturity paid in the

secondary market.4 The higher yield, hereafter called the "yield
differential," is measured as the difference between the CM bill yield
Treasury obtained at auction and the average secondary market yield on
outstanding Treasury bills that are most similar (in terms of maturity) to
the CM bill on the day of auction.5 Our statistical analysis of certain
features of CM bills, such as size issued and dates of issuance and
maturity, from fiscal year 2002-2005 (after the introduction of the 4-week
bill) shows that, all other things equal, issuing relatively larger CM
bills resulted in a higher yield differential. We also found that CM bills
that were part of a "multiple tranche"-successively issued CM bills
maturing on the same day6-tended to have higher yield differentials but
this is only significant at the 10 percent level and it is sensitive to
the inclusion of other explanatory variables. CM bills maturing around
large tax payment dates tended to have lower yield differentials. We also
found that a key driver of the yield differential is the general level of
short-term yields in the economy, a factor that is outside Treasury's
control. A continuation in the rise of short-term yields is therefore
likely to result in higher yield differentials. Although auction
performance theoretically has an effect on the yields Treasury obtained at
auction, we did not find any statistically significant relationships
between the auction performance measures we examined and the yield
differential.

Treasury can accumulate and maintain higher cash balances or issue 4-week
bills instead of using CM bills to meet short-term cash shortfalls.
However, despite their higher yield, CM bills are generally less costly
than either of these alternatives. Treasury faces a negative funding
spread-the interest earned on cash balances is generally insufficient to
cover the cost of increased borrowing necessary to maintain these
balances. The negative funding spread is due in part to the formula it
established in 1978 for the interest it earns on cash balances.7 As a
result, it is generally more cost-efficient to repay debt and then issue a
CM bill than run higher cash balances. CM bills are also less costly than
a regular 4-week bill for shorter-term needs (i.e., less than 28 days).
This is true because although 4-week bills tend to have a lower yield,
they are outstanding for a longer period of time. As a result, the
borrowing cost of a 4-week bill is generally greater than a CM bill of
shorter maturity (less than 28 days).

Treasury has taken steps to lower the use and overall cost of CM bills. In
2001, Treasury introduced the 4-week bill to help reduce the use of CM
bills. This led to a reduction in the amount of CM bills issued and their
term to maturity. Also, instead of issuing one large CM bill, Treasury has
increasingly issued CM bills in multiple tranches. Although CM bills
issued as part of a multiple tranche may have higher yield differentials
than CM bills not part of a multiple tranche, they allow Treasury to
borrow cash closer to the day it is needed, which reduces the average
issue size, term to maturity, and annualized amount outstanding.
Therefore, the savings that multiple tranche CM bills provide more than
offset the higher yields such issues might entail. Declining interest
rates, which reached historical lows in recent years, also helped reduce
borrowing costs from 2001 to 2002. However, the cost of CM bill issuance
has increased since 2003.

CM bills will continue to be a necessary debt management tool to meet
unexpected cash needs when Treasury has low cash balances and when
Treasury is nearing the debt ceiling. Our analysis indicates that the
yield differential has increased as short-term rates have risen. If these
rates rise further, as market participants expect, so will the yield
differential. While Treasury does not vary its debt management strategy in
response to changing interest rates, it should be mindful that increasing
rates are likely to raise the relative cost of unscheduled CM bills. As a
result, Treasury should consider options that may reduce the frequent use
of CM bills and ultimately overall borrowing costs.

We identified a range of options Treasury can consider that may reduce the
use of CM bills, the premium paid on CM bills, and overall borrowing
costs. One of the more promising options is exploring additional
opportunities for closer alignment of large cash flows to help eliminate
the frequent cash financing gap in the first two weeks of the month.
Treasury can also reexamine options for increasing earnings on its excess
cash balances, which would narrow the negative funding spread and may
ultimately reduce the need to rely on CM bills. Other options include
introducing a shorter-term instrument; increasing transparency on the
potential size of CM bills; relaxing rules for CM bill auctions; and
exploring other nations' debt management practices.

We recommend that Treasury explore options such as those discussed in this
report and any others it identifies that may help it achieve its
lowest-cost borrowing objective. We recognize that there are a number of
trade-offs to consider with each of these options. In its exploration
Treasury should consider the costs and benefits, including how options may
be combined to produce more beneficial outcomes, and determine whether the
benefits-in the form of lower borrowing costs-to the federal government
(and so to taxpayers) outweigh any costs imposed on individuals, business,
or other nonfederal entities. Implementing some of these options would
require changes to statutes or regulations. If Treasury determines that
any of these changes would be beneficial, we encourage Treasury to begin
discussions with relevant federal agencies and the Congress about
obtaining the necessary authorities.

In oral and written comments on a draft of this report, Treasury generally
agreed with our findings, conclusions, and recommendations. Treasury said
that it is committed to continuing to explore ways to further reduce
financing costs through changes in the use of CM bills and that many of
the options we identified are embodied in its current debt management
policy. Treasury emphasized that statutory authority is needed for some
options, particularly changing the timing of receipts and expenditures and
improving earnings on excess cash balances. Treasury also suggested some
technical changes throughout the report that we have incorporated as
appropriate. Treasury's comments appear in appendix IV. In addition, the
Federal Reserve Board provided technical comments that we incorporated as
appropriate.

Background

Congress has delegated to Treasury the power to borrow the money needed to
operate the federal government and manage the government's

outstanding debt subject to a statutory limit.8 Treasury's primary debt
management goal is to finance the government's borrowing needs at the
lowest cost over time. To meet this objective, Treasury issues debt
through auctions in a "regular and predictable" pattern across a wide
range of securities. Treasury does not "time the market"-or take advantage
of lower interest rates-when it issues securities. According to Treasury,
because investors and dealers rely upon the routine availability of
Treasury securities they tend to pay a slight premium, which lowers
Treasury's borrowing costs. In addition, Treasury also states that to
support liquidity, it must issue "enough but not too much" at each
auction.9 If Treasury issued too little, it could not sustain a deep and
liquid secondary market for its securities. If it issued too much,
Treasury creates concern among primary market participants that they may
find it difficult to distribute their holdings in the secondary market.

Treasury publishes a schedule with tentative announcement, auction, and
settlement (issue) dates up to 6 months in advance of regular security
auctions. Depending on the type of security, Treasury typically auctions
and then issues a security within a week or less. Treasury generally
issues short-term regular bills with 4-, 13-, and 26-week maturities every
Thursday and issues 2- and 5-year notes at the end of each month.10 Three-
and 10- year notes are issued in the middle of each quarter and Treasury
reopens 10-year notes 1 month after their initial issuance.11 In addition,
Treasury issues TIPS in 5-, 10-, and 20-year maturities in certain months
according to the TIPS' maturity. Finally, Treasury issues 30-year bonds in
the middle of

February12 and reopens the bonds in the middle of August. Figure 1 depicts
Treasury's April 2005 borrowing schedule.13

Figure 1: Treasury's April 2005 Regular Borrowing Schedule

Note: The April 2-year note was issued on May 2 since the end of April
fell on a weekend. The 3-year note was not issued since April did not fall
in the middle of the quarter (i.e., quarterly refunding months-February,
May, August, and November).

Treasury supplements its regular and predictable schedule with flexible
securities called CM bills. Unlike other securities, Treasury does not
publish information on CM bills on its auction schedule. Instead, Treasury
generally announces CM bill auctions anywhere from 1 to 4 days ahead of
the auction. The term to maturity-or length of time the bill is
outstanding-varies according to Treasury's cash needs. CM bills allow
Treasury to finance very short-term cash needs-for as little as 1
day-while providing short notice to market participants.

The United States Treasury is not alone in using CM bill-type instruments
to finance short-term needs and smooth cash flows. For example, in the
United Kingdom, the Debt Management Office (DMO) issues CM bills to meet
temporary cash flow needs that the DMO cannot conveniently meet through
its structured bill auctions and to help smooth cash flows. The Bank of
Canada, which auctions securities for Canada's debt management, issues CM
bills to help it minimize the level and cost of carrying cash balances.
Australia uses short-term securities with 5-, 13-, and 26-week maturities
to bridge within-year cash flow mismatches, but may vary the maturities
and issue shorter-term securities of 3 to 6 weeks.

Objectives, Scope, and Methodology

To understand when Treasury uses CM bills, we analyzed CM bills issued
over the last 10 fiscal years-fiscal years 1996-2005. This time period
provides a sufficiently large sample of CM bills-121-and allowed us to
analyze and show trends in CM bill use before, during, and after the
following events: 4 years of surpluses, five debt issuance suspension
periods (DISP) declared by the Secretary of the Treasury,14 and the
introduction of two new debt instruments-TIPS in 1997 and the 4-week bill
in 2001. CM bill data were obtained from the Bureau of the Public Debt
(BPD). BPD's online database has 22 different features for each Treasury
security, including the amount, announcement, auction and issue dates, the
maturity date, and the yield. We examined these features as well as others
examined in earlier studies of CM bills to identify patterns of CM bill
auctions, issuance, and maturity.

To understand why CM bills are issued or mature at these times, we
examined Treasury's cash flows from fiscal years 1996-2005 using publicly
available data from the Financial Management Service's (FMS) Daily
Treasury Statements. We also met with Treasury officials from the Office
of Debt Management and the Office of Fiscal Projections. We reviewed
Treasury documents, including quarterly refunding statements and policy
statements, and Treasury Borrowing Advisory Committee reports and minutes
from quarterly refunding meetings.15

We obtained views on the perceived advantages and disadvantages of CM
bills for Treasury, Federal Reserve operations, and investors in meetings
with Treasury officials, Federal Reserve officials, market participants,
including primary dealers and money market fund managers, and market
analysts. We also reviewed financial and economic literature. We did not
identify any key advantages or disadvantages of CM bills for Federal
Reserve operations and as a result these operations are not a focus in
this report.

To describe the key disadvantage of CM bills for Treasury-higher yields-we
estimated the differential between CM bill yields and the yields on
outstanding Treasury bills of similar maturity at the time of auction
using data from BPD and the Wall Street Journal (WSJ).16 To determine
whether certain features might reduce the yield paid on CM bills, we
regressed the yield differential on key features. We determined which
features to examine on the basis of previous studies of CM bills and
Treasury auctions, our interviews with Treasury and market participants,
and our own analysis. While we analyzed the yield differential for all CM
bills issued during fiscal years 1996-2005, we focused on the 55 CM bills
issued during fiscal years 2002-2005 because the introduction of the
4-week bill in 2001 substantially reduced CM bill maturities and caused a
structural change in the CM bill market. For more information on our
statistical analysis, see appendix I.

Because CM bill rates are determined in auctions, we examined auction
performance to determine whether there is any relationship between
selected performance measures and the yield differential. We also examined
whether CM bill auctions perform as well as regular 4-week bill auctions
because the 4-week bill is more similar to a CM bill-in terms of issuance
amount variance and its term to maturity-than other Treasury securities.
We evaluated each of the 284 CM and 4-week bill auctions in fiscal years
2001-2005 (years since the introduction of the 4-week bill) using
proprietary data from the Treasury Auction Database (TAD) and GovPX-an
interdealer broker database with information on primary dealer
transactions for all U.S. Treasury securities. We analyzed auction
performance measures mentioned in Treasury interviews, Treasury studies of
auctions,17 Borrowing Advisory Committee minutes, and economic and
financial literature/textbooks describing or evaluating Treasury auction
performance. See appendix II for more information on our analysis of CM
bill auctions.

We identified possible options to reduce the use and cost of CM bills on
the basis of our analysis of CM bill use, the yield differential, and CM
bill auction performance. We discussed these options with Treasury, market
participants, and others and we include their comments as appropriate.

In order to assess the reliability of data used in this study, including
proprietary data from TAD and GovPX, and the publicly available data from
FMS, BPD, and WSJ, we examined the data to look for outliers and anomalies
and addressed such issues as appropriate. In general, we chose databases
that were used by Treasury and researchers to examine Treasury markets and
auction performance. Where possible and appropriate, we corroborated the
results of our data analysis with other sources. On the basis of our
assessment we believe the data are reliable for the purposes of this
review.

We conducted our review in Washington, D.C., from February 2005 through
March 2006 in accordance with generally accepted government auditing
standards.

Treasury Uses CM Bills to Help Fill Cash Financing Gaps

Treasury frequently faces a cash financing gap of about 2-weeks duration
because of timing differences of large cash inflows and outflows. Treasury
makes large regular payments (e.g., Social Security and federal
retirement) in the beginning of the month and it often receives large cash
inflows in the middle of the month from income tax payments and note
issuances. Because regular bills alone are not sufficient to fill this
cash financing gap, Treasury has increasingly used CM bills since 2002.
Treasury has also relied on CM bills when nearing the debt ceiling to help
pay its bills while keeping debt under the statutory limit.

Large Cash Outflows Occur at the Beginning of the Month

Treasury's largest cash outflows generally occur in the beginning of the
month. For example, in fiscal year 2005, almost one-quarter of the
government's annual fiscal cash outlays (withdrawals excluding debt
redemption) were paid in the first 3 days of each month.18 On or around
the 1st of every month Treasury paid about $14 billion to active duty
military personnel, military and civilian retirees, and others. In the
beginning of some months, Treasury also paid up to $6 billion for
Medicare. On or around the 3rd of every month it paid about $21 billion in
Social Security benefits. In total, Treasury made $718 billion cash
payments in the first 3 days of months in fiscal year 2005.

These large payments in the beginning of the month are not anticipated to
decline soon. A Treasury official explained that Social Security benefits
paid on the 3rd of the month are anticipated to remain relatively steady
for a number of years and then decline because of steps taken by the
Social Security Administration (SSA) in 1997 that have helped smooth cash
payments out of Treasury. (Fig. 2 describes these steps in more detail.)
However, because beneficiaries receiving benefits before 1997 continue to
receive their benefits at the beginning of the month, we estimate the
large payments could last another 10 years. Other federal benefit programs
continue to pay all or most of their benefits at the beginning of the
month. If these payments were, like Social Security, spread throughout the
month, it would help smooth cash flows and might reduce Treasury's need to
use unscheduled large CM bills at the beginning of the month.

Figure 2: Smoothing the Payment of Social Security Benefits

Large Receipts Often Occur in the Middle of the Month

In contrast to outlays, Treasury's largest cash inflows generally occur in
the middle of the month. Although the majority of federal tax receipts are
already fairly smoothed throughout the year, most corporate and
nonwithheld individual tax payments are made in the middle of certain

months throughout the year.19 Treasury receives large corporate tax
payments on (or around) the 15th of March, April, June, September, and
December. During fiscal year 2005, Treasury received just over $213
billion of cash on these days. Treasury also had large receipts in the
middle of January, April, June, and September from nonwithheld individual
estimated tax payments, and after April 15 from the settlement of prior
year individual income tax liability.

In addition to tax receipts, net cash raised from note issuances (i.e.,
issuance less redemption) on or around the 15th of the month more than
tripled from $78 billion in fiscal year 2002 to $297 billion in fiscal
year 2005. However, some large midmonth cash inflows from note issuance
will not endure because Treasury shifted 5-year note auctions from the
middle to the end of each month beginning in February 2006. As a result,
cash inflows from new 5-year debt issuance shifted from the middle to the
end of the month.

Regular Bill Issuances Alone Are Not Sufficient to Fill Cash Financing Gap

To achieve the lowest borrowing costs over time, Treasury seeks to provide
the market with a high degree of stability in the amount issued of each
security, with the requisite stability increasing with issuance maturity.
Treasury officials said they try to limit swings in regular bill
offerings. This means regular bill issuances cannot suddenly increase by
the amount needed to make large payments in the beginning of the month nor
suddenly decrease in the middle of the month to absorb large inflows. In
practice, Treasury only varied the amount of 13- and 26-week bills by $2
billion (less than 10 percent of the average amount issued) from week to
week in fiscal year 2005. Since Treasury introduced the 4-week bill in
2001 to help reduce cash balance swings, its issuance size varies more
than that of other regular Treasury bills. In fiscal year 2005, the size
of the 4-week bill varied by as much as $13 billion (more than
three-quarters of the average amount issued). Even so, to meet cash
needs-which in fiscal year 2005 averaged almost $60 billion at the
beginning of the month-Treasury has come to rely on CM bills.

Treasury Has Increasingly Used CM Bills to Fill Frequent Cash Financing
Gaps

The combination of low cash balances in the beginning of the month with
large cash inflows in the middle of the month has led to a general pattern
of CM bill issuance. Our analysis shows that from 2002 to 2005, on average
about half of CM bills were issued in the first 3 days of the month.
According to Treasury officials, Treasury decides to issue CM bills when
it has low cash balances, which often track the timing of large payments.
In fiscal year 2005, for example, Treasury issued 11 of its 21 CM bills in
the first 3 days of the month when it had below average cash balances (the
average cash balance for fiscal year 2005 was $25.5 billion) (see fig. 3).
The amounts issued varied from $4 billion to $42 billion.

Figure 3: Cash Balance and CM Bill Issuance during Fiscal Year 2005

The maturity dates of CM bills have varied over the 10-year period we
examined, but since 2002 CM bills have increasingly matured on the 15th of
the month when Treasury receives large cash inflows. In fiscal year 2005,
Treasury set 68 percent of its $268 billion CM bill borrowings to mature
in the middle of March, April, June, September, and December when Treasury
receives large corporate and individual income taxes. In these 5 months
Treasury received 72 percent ($213.3 billion) of fiscal year 2005
corporate income tax deposits.20 With only one exception, all CM bills
issued in fiscal year 2005, regardless of when Treasury issued them,
matured on or around the 15th of the month (see fig. 4).21

Figure 4: Cash Receipts and Maturity Dates of CM Bills Issued during
Fiscal Year 2005

aCash receipts exclude net issuance of debt.

Since 2002, Treasury has increasingly filled the approximately 2-week
financing gap by issuing CM bills at the beginning of the month and
setting

the maturity date of these CM bills for the middle of month.22 Table 1
shows that since 2002, 23 CM bills-or more than 40 percent of the number
of CM bills issued-have been issued on the 1st-3rd days of the month and
matured on the 15th day of the month. These CM bills accounted for more
than half of the total dollar amount of CM bills issued in the last 4
fiscal years.

Table 1: CM Bills Issued on the 1st-3rd Days of the Month That Mature on
the 15th Day of the Month

Number Percent of total Amount (billions Percent of
number issued of dollars) total amount
issued
1996 0 0% $0 0%
1997 1 11 26 16
1998 1 14 15 9
1999 5 45 99 42
2000 1 10 25 11
2001 5 42 162 47
2002 0 0 0 0
2003 3 25 69 39
2004 9 60 202 82
2005 11 52 175 65
Total (1996-2001) 13 20% $327 23%
Total (2002-2005) 23 42% $446 55%

Source: GAO analysis of BPD data.

Treasury Also Uses CM Bills to Help Smooth Cash Flows during the April Tax
Season

Over the last 10 years Treasury has relied on multiple CM bills to help
manage cash flows in April. The large cash balance swings in April could
not be accommodated by changes in the regular bill issuance schedule
according to Treasury officials and our analysis. To help smooth large
April cash flows, Treasury consistently issued three or more CM bills
every year between fiscal years 1996 and 2005 that matured in the later
half of April (on the 15th or later), when Treasury typically received
large tax receipts. For example, in fiscal year 2005 Treasury issued three
CM bills-two matured on April 15 and one matured on April 18-totaling $47
billion and received about $41 billion in corporate and nonwithheld income
and employment tax payments on those dates.

Treasury Does Not Control the Timing of All Cash Flows

Better aligning or smoothing cash inflows and outflows could reduce or
eliminate the frequent financing gap thereby reducing the need for some CM
bills but Treasury does not have authority to control the timing of all
cash flows. Treasury is a passive agent; it collects and disburses federal
funds at agencies' request.23 It does not determine when major benefit
payments are made. For example, the payment dates of civil service and
railroad retirement are set by law. Due dates for tax payments are also
set by federal statute.24

Treasury does have some control over debt-related cash flows and has
considered how changes in its schedule affect cash flows. For example,
when making changes to its schedule in 2001, Treasury maintained monthly
issuance of 2-year notes, which are issued at the end of the month and
could help pay regular benefit payments in the beginning of the following
month. More recently, Treasury moved 5-year note issuance from the middle
to the end of the month. This change more closely aligns 5-year note
issuance with the beginning of the next month's payments and may reduce CM
bill issuance in the early part of months.

However, regular securities can only help fill cash financing gaps
temporarily. No regular security that matures on the same day another is
issued can be used to fill cash financing gaps over the long run because,
in a steady state, most, if not all, of the cash raised at each issuance
would be needed to pay maturing securities. As a result, Treasury would
not raise large amounts of new cash. In addition, Treasury generally sets
the issue amount for longer-term securities to cover long-term deficit
needs, not short-term cash shortfalls, and thus the new cash raised with
the 5-year note may not cover the full amount of cash needs in the
beginning of the month. Although Treasury will likely raise new cash at
the end of the month with the 5-year note until 2011 when 5-year notes
issued in 2006 begin to mature, the change in the auction schedule is not
likely to eliminate the need for CM bills.

Treasury Also Relies on CM Bills When Approaching the Debt Ceiling

When debt is nearing the statutory limit, Treasury has to take a number of
extraordinary steps to meet the government's obligation to pay its bills
while keeping debt under the ceiling.25 Treasury also issues CM bills,
among other actions, to accomplish these goals. CM bills, like other
Treasury securities, are subject to the debt limit. However, CM bills
allow Treasury to borrow cash for shorter time periods than regular bills.
On occasion, Treasury has changed scheduled auctions for its regular
securities and instead issued CM bills so that the debt ceiling will not
be reached. For example, because of debt limit constraints Treasury
delayed the 4-week bill auction scheduled for Tuesday, November 16, 2004.
Treasury then auctioned a 5-day CM bill for $7 billion on November 17,
2004. In another example, Treasury said that inaction on the debt ceiling
in 2002 led to reduced issuance of 4-week bills and larger, more frequent
CM bill issuance than it would have done otherwise. During the five DISPs
within the last decade, Treasury issued 19 CM bills totaling $300 billion.
As table 2 shows, Treasury issued most of these CM bills during the
lengthy DISPs in fiscal years 1996 and 2003.26

Table 2: DISPs over the Last Decade (Fiscal Years 1996-2005)

Fiscal DISP beginning DISP ending Length Number of Amount (billions of
year date date (days) CM bills dollars)
issued
1996 November 15, March 29, 135 8 $130
1995 1996
2002 April 4, 2002 April 16, 12 2 39
2002
2002 May 16, 2002 June 28, 2002 43 2 41
2003 February 20, May 27, 2003 96 6 83
2003
2005 October 14, November 19, 36 1 7
2004 2004
Total 19 $300

Source: GAO analysis of BPD data.

CM bills will continue to be a useful tool for Treasury when it approaches
the debt ceiling, but, as discussed on page 20 of this report, Treasury
pays a premium for the flexibility CM bills provide.

CM Bills Cannot Be Used in All Circumstances and Other Alternatives No
Longer Exist or Are Limited

Although CM bills offer a means to raise cash in as little as a day, CM
bills, like other Treasury securities, cannot be used when financial
markets are closed or not functioning properly. Previously, Treasury could
obtain cash on short notice outside financial markets, but this capability
ceased more than twenty years ago. Treasury was also able to draw on
compensating balances when financial markets closed after September 11,
2001, but these balances were terminated in 2004.

In the past, Treasury had access to a cash draw authority. Intermittently
between 1942 and 1981, Treasury was able to directly sell (and purchase)
certain short-term obligations from the Federal Reserve in exchange for
cash. Treasury used the cash draw authority infrequently and mostly in
times of war or armed conflict. The Federal Reserve held special
short-term certificates purchased directly from Treasury on 228 days
between 1942 and 1981. In the years Treasury used this authority, it
borrowed on about 11 days on average per year. The Treasury Draw Policy,
as amended in 1979, stated that Treasury could use the cash draw authority
only in

"unusual and exigent circumstances."27 Congress allowed this authority to
expire in 1981.28

Prior to March 2004 Treasury could use compensating balances-noninterest
bearing cash balances that were used to compensate banks for various
services-as a source of short-term funding when markets were closed or
during DISPs. Treasury officials said that compensating balances were not
viewed as a substitute backup facility for Treasury to obtain cash in the
short term and were only used in extraordinary circumstances. For example,
on September 11, 2001, Treasury had to cancel the auction of 4-week bills,
which would have settled on Thursday, September 13, 2001. Because of the
auction cancellation, Treasury lacked sufficient cash to pay about $11
billion of maturing 4-week bills on Thursday, September 13. Treasury
obtained sufficient cash by drawing down compensating cash balances.
However, in March 2004, compensating balances were replaced with direct
payments to the banks.

More recently, in the aftermath of Hurricane Katrina, when cash balances
fell more than expected, Treasury obtained cash by canceling a planned
cash investment. Treasury periodically auctions excess cash to banks
through its Term Investment Option (TIO) program. Treasury invests cash
through TIOs for a fixed term at a rate determined through a competitive
auction process.29 Treasury intended to award a $5.5 billion investment
option on Friday, November 25, 2006. However, additional spending in
response to Hurricane Katrina caused Treasury's cash balance to fall to an
unexpected low level on November 23, 2005. In response, Treasury did not
follow through on the planned investment option.

Using these investments as a source of cash in the future requires that
Treasury actually have excess cash to invest and that it knows it will
need the cash before investment. Reliance on excess cash is limited since
Treasury has placed increased emphasis on minimizing cash balances in
order to reduce overall borrowing costs. One possible consequence of this
practice is an increased risk that incorrect cash flow predictions or
emergencies could lead to Treasury overdrawing its Federal Reserve Bank
account. As a result, an important issue for future consideration is how
Treasury might obtain funds to finance government operations should normal
financial market operations be degraded significantly or closed because of
a catastrophic emergency.

CM Bills Generally Had Higher Yields than Outstanding Bills of Similar
Maturity but Lower Borrowing Costs than Currently Available Alternatives

CM bills provide Treasury with flexibility to obtain cash outside its
regular borrowing schedule, but Treasury generally paid a higher yield on
CM bills than outstanding bills of similar maturity paid in the secondary
market. The differential between CM bill yields and similar maturing
outstanding bills-hereafter called the "yield differential"-varied greatly
from fiscal year 1996 through 2005.30 We found several factors, both
within and outside Treasury's control that affected the yield
differential. Despite their higher yield, CM bills are generally less
costly than maintaining higher cash balances or issuing 4-week bills as
means to obtain the cash needed to make large payments, such as Social
Security and federal retirement, at the beginning of the month.

CM Bill Yields Closely Track Other Short-Term Rates

CM bill yields closely track the level of short-term interest rates
prevailing at the time of auction. This is because investors see other
short-term instruments as close substitutes for CM bills. They can either
buy new bills or buy existing bills in the secondary market. In general,
investors will not accept a lower yield (bid a higher price) for a new
bill than that available on an existing bill. Conversely, investors would
not offer a lower price to obtain a higher yield than prevailing in the
market since they would likely be underbid. As a result, CM bill yields
follow other short-term yields within a narrow range.

Short-term yields change over time in response to changes in economic
activity, the demand for credit, investors' expectations, and monetary
policy as set by the Federal Reserve. CM bill yields declined from about
5.8 percent in 2000 to 1.1 percent in 2004. This decline reflected the
overall reduction in short-term rates driven largely by the Federal
Reserve's monetary actions and other market forces. The Federal Reserve
started lowering the federal funds rate-the interest rate at which banks
lend reserves to other banks overnight-in early 2001, and by 2002 the
federal funds rate was at levels not seen since the early 1960s. Beginning
in the summer of 2004, the Federal Reserve began to increase the federal
funds rate. At the same time, CM bill yields increased to an average 2.5
percent in fiscal year 2005 (see fig. 5).

Figure 5: Annualized Federal Funds Rate and CM Bill Yields (Fiscal Years
1996-2005)

aAnnualized number based on average of federal funds rates on days of CM
bill auctions. CM bill rates converted to money market yields for
comparability.

CM Bill Yields Were Generally Higher than Yields on Outstanding Bills of
Similar Maturity

Treasury paid a higher yield on most CM bills issued during our study
period than outstanding bills of similar maturity paid in the secondary
market. The average yield differential fell from 47 basis points31 in
fiscal year 2001 to 5 basis points in fiscal year 2004 (see fig. 6). In
fiscal year 2005, yield differentials grew and CM bill yields were about
14 basis points higher on average than outstanding bills of similar
maturity.

Figure 6: Yield Differential (Fiscal Years 1996-2005)

Note: The yield differential compares CM bill yields with the weighted
average of yields on outstanding Treasury bills maturing before and after
the CM bill. For more information, see app. I, pp. 53-54.

Our analysis identified two important factors behind the yield
differential decline: lower short-term Treasury yields and reduced CM bill
issuance. The first effect was somewhat temporary while the latter effect
could last. The level of short-term interest rates is largely driven by
Federal Reserve policy and market forces rather than by Treasury. Treasury
bill yields have risen and may continue to rise and eventually reach
levels that prevailed in the late 1990s, thereby erasing the portion of
the decline in the yield differential caused by lower interest rates.
However, because the 4-week bill is now a permanent feature of Treasury's
auction schedule and has reduced Treasury's reliance on CM bills, the
portion of the decline in the yield differential attributable to
relatively lower CM bill issuance is likely to endure. The experience
during fiscal years 2002-2005 as a whole suggests that the yield
differential could remain about 13 basis points below pre-2002 levels.
These findings are discussed in more detail later in this report.

The large reduction in the yield differential has helped reduce borrowing
costs associated with CM bills. The daily cost associated with the
47-basis-point yield differential in fiscal year 2001 was about $12,900
per $1 billion. In fiscal year 2001, Treasury borrowed $19.2 billion
(annualized amount outstanding) using CM bills.32 Of the $1.06 billion in
total borrowing costs associated with CM bills in that year, $70 million
was associated with the yield differential. Since then, the average yield
differential has declined and was 14 basis points, or about $3,800 a day
per $1 billion borrowed, in fiscal year 2005. During fiscal year 2005,
Treasury borrowed about $8 billion (annualized amount outstanding) using
CM bills. Total borrowing costs were $215 million and the borrowing cost
associated with the yield differential was about $12.8 million. Treasury
could achieve savings by further reductions in the yield differential.33

CM Bill Yields Were Similar to Yields on Overnight Repurchase Agreements

CM bills may have higher yields because, according to Treasury officials
and market participants, they are bought for a different purpose than
regular bills. According to market participants, some money market funds
and foreign central banks purchase CM bills but for the most part they are
not widely used as an investment tool because of their irregularity and
short-term nature. Instead, CM bills are primarily used by primary dealers
as collateral for repurchase agreements. A repurchase agreement is a form
of short-term collateralized borrowing used by dealers in government
securities. Figure 7 describes repurchase agreements in more detail.

Figure 7: Repurchase Agreements

aAgency securities are financing instruments of federal government
agencies, such as the Tennessee Valley Authority, or government-sponsored
enterprises, such as the Federal National Mortgage Association (Fannie
Mae) and the Federal Home Loan Mortgage Corporation (Freddie Mac). Agency
securities are not considered risk free but are considered to have high
credit quality in part because of the agencies' government affiliation.

We found the yields of CM bills to be near the yields of overnight
repurchase agreements. In fiscal year 2005, CM bill yields were within 2
basis points of overnight repurchase agreement yields.34 Although
repurchase agreements and CM bills are both short-term investments, there
are some differences. For example, repurchase agreements are subject to
federal, state, and local taxes whereas CM bills are exempt from state and
local taxation. Also, CM bills, like regular Treasury bills, are risk free
whereas repurchase agreements issued by private borrowers involve some
risk. The high-quality collateral in a repurchase agreement (e.g.,
Treasury securities or agency securities) reduces the credit risk faced by
the lender and allows borrowers to obtain cash at a lower cost than they
would obtain otherwise.35 However, the lender is still exposed to credit
risk because if the borrower fails to repay the loan, the market value of
the collateral may be less than the amount owed. Repurchase agreements are
structured carefully to reduce this credit risk exposure, for example, by
lending less than market value of the security used as collateral.
Nevertheless, repurchase agreements have more risk than securities issued
by Treasury and accordingly have higher yields than Treasury securities of
similar maturity, which are risk free and are the floor for short-term
rates in the money market.

Our Analysis Identified Some Features That Affect the Yield Differential

Our statistical analysis found that, all other things equal, the amount of
CM bills issued was positively correlated with the yield differential,
meaning that increases in the amount of CM bills issued relative to the
amount of similar-maturing Treasury bills outstanding increased the yield
differential, and decreases in the amount issued reduced the yield
differential.36 We also found that CM bills that were part of a "multiple
tranche"-successively issued CM bills maturing on the same day-tended to
have higher yield differentials,37 and CM bills maturing around large tax
payment dates tended to have lower yield differentials. Changes in the
general level of short-term yields also affected the yield differential.
While the amount and timing of CM bills is somewhat under Treasury's
control, the general level of short-term yields in the economy is not.38

Although CM bill yields are determined in auctions, we did not find a
statistically significant relationship between the auction performance
measures we examined-such as participation, distribution of auction
awards, and preauction activity in the when-issued market39-and the yield
differential. However, we found that in more concentrated auctions
Treasury was less likely to pay more than the yield indicated in the
when-issued market at the time of auction, which is a positive auction
result.40

The Amount of CM Bills Issued Was Positively Correlated with the Yield
Differential

While an increase in the relative amount of CM bills might enhance
liquidity and thereby reduce their yields, investors might require a
higher yield to acquire relatively large amounts of CM bills. Our
statistical analysis found that increases in the amount of CM bills issued
relative to outstanding bills with similar maturity increased the yield
differential. For example, if Treasury had increased its fiscal year 2005
CM bill issuance by $10 billion, the average yield differential would have
been about 4.2 basis points higher according to our analysis. Conversely,
reducing the amount of CM bills issued tended to reduce the yield
differential. Earlier research by others had similar findings.41

According to our analysis, the reduced use of CM bills since the 4-week
bill was introduced contributed to the overall decline in the yield
differential. From fiscal years 1996-2001, the amount of CM bills issued
was almost 70 percent of outstanding bills with similar maturity. In
fiscal year 2002, the share declined to 30 percent. At the same time, the
yield differential declined from about 35 basis points (fiscal year
1996-2001 average) to 4 basis points in 2002. According to our analysis,
about 35 percent of the yield differential decline can be attributed to
Treasury's reduced use of CM bills. CM bill issuance has remained low
relative to outstanding bills with similar maturity since 2002. These
results suggest that the 4-week bill reduced both the use and cost of CM
bills as Treasury intended.

Issuing CM Bills in Multiple Tranches May Increase the Yield Differential
but Reduce Overall Borrowing Costs

We also found that consecutive CM bills maturing on the same day-called
"multiple tranches"-tended to increase the yield differential. In recent
years, Treasury has increasingly issued consecutive, small CM bills with
the same maturity date-rather than one larger CM bill-which reduces the
average amount outstanding and ultimately reduces borrowing costs. In
2005, three-quarters of the CM bills issued were part of a multiple
tranche. Our analysis suggests that issuing consecutive CM bills that
mature on the same date might increase the yield differential by 6 basis
points over a single CM bill. Market participants suggested if they
suspect Treasury might reopen the CM bill later, they may bid less
aggressively on a CM bill, which would result in lower prices and higher
yields for CM bills.

The evidence that issuing in multiple tranches increases the yield
differential should be viewed with caution. The estimated coefficient is
only significant at the 10 percent level and is sensitive to the inclusion
of other explanatory variables.42 For example, with the inclusion of a
variable representing the number of days advance notice before auction,
the coefficient of multiple tranche CM bills is no longer significant even
at the 10 percent level. Moreover, the lower cost of multiple tranches
resulting from smaller average issues and shorter term to maturities most
likely offsets any increase in the yield. For more information on the
cost-saving effects of multiple tranches, see pages 38-39.

CM Bills Maturing on Large Tax Payment Dates May Have Lower Yield
Differentials

Increasingly Treasury has issued CM bills that mature on individual and
corporate tax payment dates and, according to our analysis, this practice
may be leading to lower yields and borrowing costs. Our statistical
analysis shows the yield differential on CM bills maturing on or near
April 15 or on other tax payment dates (generally the 15th of March, June,
September, and December) is 12 basis points lower than other CM bills.
This may be explained by empirical evidence suggesting that regular
Treasury bills whose maturity dates immediately precede corporate tax
payment dates have special value because corporate treasurers may wish to
invest excess cash in securities whose cash flows can be used to liquidate
cash liabilities.43

Almost 50 percent (27 of 55) of CM bills Treasury issued over the last 4
fiscal years matured on large tax payment dates. In the last 2 fiscal
years, Treasury set at least one CM bill to mature on each of the large
tax payment dates. So, to a large extent, Treasury has already captured
the cost-savings from this feature. It is also important to note, however,
that the 12-basis point difference is unlikely to offset even an extra day
of borrowing. CM bills maturing on dates other than large tax payment
dates with shorter maturities are likely to cost less than CM bills
maturing on a corporate tax payment date with longer maturities because
debt is outstanding for a shorter period of time.

Yield Differential Rises and Falls with Short-term Yields

Our analysis showed that the yield differential rises and falls with the
overall level of Treasury bill yields. Figure 8 shows short-term Treasury
yields declined from an average of about 4.8 percent in fiscal year 2001
to 1.05 percent in fiscal year 2004. At the same time, the average yield
differential fell from 47 basis points to 5 basis points. Our analysis
suggests the decline in short-term yields explained about 35 percent of
the decline. Since mid-2004 when short-term rates started to increase
consistent with Federal Reserve actions, the yield differential has
started to widen again to about 14 basis points in fiscal year 2005. Thus,
absent Treasury actions, further increases in short-term rates are likely
to lead to higher CM bill yield differentials in the future. If short-term
Treasury yields return to the 1996-2001 average of approximately 5
percent, we found that the yield differential could exceed 20 basis points
and the additional cost of CM bills-assuming current issuance
patterns-could increase from $11 million to about $19 million a year.

Figure 8: Yield Differential and the Average Level of Short-term Treasury
Yields (Fiscal Years 2001-2005)

CM Bills Are Innately Less Predictable than Regular Bills

The yield differential may also be attributable in part to variables that
are hard to measure, such as predictability. According to Treasury
officials, a regular and predictable borrowing schedule is attractive to
investors and helps to achieve Treasury's objective of lower borrowing
costs over time. However, both the timing and amount of CM bill auctions
are by nature less predictable than regular Treasury bills. Providing too
much advanced notice of CM bill issuance would reduce Treasury's
flexibility to adjust the amount and timing of issuance to best meet cash
needs. It could also cause Treasury to use conservative estimates of its
cash needs and borrow more than it actually needs. Treasury's challenge is
to provide market participants enough notice to avoid paying too high a
premium for uncertainty without reducing its flexibility. Improving cash
forecasting could help Treasury determine the amount of CM bills to be
issued sooner rather than later and thus provide market participants more
notice of CM bill auctions.

Treasury provides limited information on the timing and amount of CM bill
auctions ahead of the announcement. The Quarterly Refunding process is
Treasury's way to provide market participants with information and get
their feedback about changes to its auction schedule and the issues
actively under discussion by Treasury. As part of each Quarterly
Refunding, Treasury usually issues a statement indicating that it plans to
issue CM bills in the coming quarter. Treasury provides general
information on the timing of CM bills, such as "early March" or "early
April" but does not provide the actual date it expects to issue CM
bills.44 In contrast, Treasury publishes the auction schedule, including
announcement, auction, and issue dates for its regular bills up to 6
months in advance. Treasury typically does not provide an estimated issue
amount for CM bills or regular bills in its Quarterly Refunding
statements.

Despite the general notice given by Treasury and the somewhat regular
pattern of CM bill issuance in recent years, market participants cannot
always predict the timing of CM bill auctions with certainty. For example,
in fiscal year 2005, a prominent money market analyst predicted the exact
date of 6 (out of 21) CM bill auctions at least 1 week ahead of the
auction announcement and was off by 1 or 2 days for 9 CM bill auctions.
However, the analyst's predicted dates for 2 CM bill auctions were off by
6 days, and the analyst did not predict the remaining four CM bill
auctions.

Not only the timing, but also the amount of CM bill issues cannot be
predicted with certainty. Treasury uses CM bills to manage cash balance
swings and as a result, the amount of CM bills issued varies more than
regular and predictable securities. Thus, it is not surprising that market
participants cannot predict with certainty the amount of CM bill issuance.
Issuing in multiple tranches may exacerbate the issue. For example, one
market analyst predicted Treasury would auction one CM bill in the first
week of December 2004 for $28 billion and one CM bill the following week
for $7 billion. Instead, Treasury auctioned three CM bills totaling $42
billion in the first week and none the following week. Several market
participants we spoke with said that uncertainty surrounding the amount
offered affects their bidding and ultimately Treasury's borrowing terms.
If market participants are uncertain of the amount, they may not bid as
aggressively, which potentially reduces the price and increases the yield
Treasury pays on CM bills. In addition, the relative inflexibility of
Treasury's demand for cash to avoid a negative cash balance might explain
the higher yields on CM bills.45

Most Measures We Examined Suggest CM Bill Auctions Perform Less Well Than
4-week Bill Auctions

Given that CM bill auctions are less predictable than regular Treasury
bill auctions, it is not surprising that CM bill auctions do not perform
as well as regular bill auctions by some measures. Better auction
performance can be characterized by greater participation and more
preauction activity in the when-issued market. These factors theoretically
reduce Treasury's borrowing costs. By most measures of participation and
activity we examined, CM bill auctions perform less well than 4-week bill
auctions. However, some other measures indicate that Treasury obtained a
better price at CM bill auctions compared with 4-week bill auctions and
that there is stronger demand for CM bills than 4-week bills.

In order to lower borrowing costs, Treasury seeks to encourage more
participation in auctions. In general, large, well-attended auctions
improve competition and lead to lower borrowing costs for Treasury. We
found that fewer bidders in total are awarded CM bills than 4-week bills.
For example, of the 55 CM bill auctions held between 2002 and 2005, more
than half had 16 or fewer awarded bidders. In contrast, only about 4
percent of the 209 4-week bill auctions had 16 or fewer awarded bidders
while about 50 percent had at least 22 awarded bidders.

We also found that preauction trading activity is sparse before CM bill
auctions. Treasury auctions are preceded by forward trading in markets
known as "when-issued" markets. The when-issued market is important
because it serves as a price discovery mechanism that potential
competitive bidders look to as they set their bids for an auction.
When-issued trading reduces uncertainty about bidding levels surrounding
auctions and also enables dealers to sell securities to their customers in
advance of the auction so they are better able to distribute the
securities and bid more aggressively, which results in lower costs to
Treasury. We counted the number of preauction trades on the day of CM bill
auctions and found that when-issued trading is lower prior to CM bill
auctions than regular 4-week bill auctions.

A broad range of participants generally improves competition and
theoretically maximizes the price investors pay for Treasury securities.
Alternatively a higher concentration-a large share of the auction awarded
to few participants-could reduce competition and restrict a security's
supply in the secondary market, preventing its efficient allocation among
investors. We evaluated the share of the auction awarded to the top five
bidders-a measure used by Treasury in its own studies of auction
performance-and found that the share was 60 percent or higher in over half
(34 of 55) of CM bill auctions from fiscal year 2002 through 2005. In
contrast, the share exceeded 60 percent in only 18 percent (37 of 209) of
4-week bill auctions in the same period. While a higher concentration
theoretically reduces competition and the price investors pay, according
to Treasury a high concentration ratio in CM bill auctions may imply some
bidders really want a particular bill, which may drive the price up and
yield down.

Although greater participation, a broader distribution of awards, and more
preauction activity in the when-issued market theoretically improves
Treasury's borrowing costs, we did not find a statistically significant
relationship between these factors and the yield differential. However, we
did find that concentration was negatively correlated with the auction
spread. The auction spread is the difference between the yield Treasury
obtains at auction and the yield in the when-issued market at the time of
auction. A positive spread (where the auction yield is more than the
contemporaneous when-issued yield) indicates Treasury paid a higher yield
than expected, which is a negative auction result. Our statistical
analysis suggests that Treasury was less likely to pay more than the
expected yield indicated in the when-issued market in more concentrated
auctions. In other words, higher concentration seemed to improve
Treasury's auction results. App. II provides more information on our CM
bill auctions analysis.

CM Bills Are Generally Less Costly than Currently Available Alternatives

Treasury could maintain higher cash balances or issue regular Treasury
bills (e.g., the 4-week bill) to avoid issuing a CM bill to meet
short-term cash shortfalls. When evaluating CM bills relative to other
alternatives, it is important to look at total borrowing costs-not just
the yield. Treasury's borrowing costs are based on the amount borrowed,
the yield it pays to borrow, and the time the debt is outstanding. We
found that CM bills are generally less costly than currently available
alternatives despite their higher yield.

CM Bills Are Generally Less Costly than Running Higher Cash Balances

To avoid issuing a CM bill, Treasury could run higher cash balances to
bridge cash financing gaps. However, it is generally more cost-efficient
to repay debt and then issue a CM bill than run higher cash balances
because the interest earned on excess cash balances is generally
insufficient to cover borrowing costs. Treasury's current cash balance
target is $5 billion, which represents the amount to be held at the
Federal Reserve. Treasury invests excess cash above the $5 billion target
in Treasury Tax and Loan (TT&L) accounts.46 TT&L accounts are held at
financial institutions and earn interest rates equal to the federal funds
rate less 25 basis points.47 The rate earned on TT&L accounts is generally
less than the average rate Treasury pays on CM and regular short-term
bills. As a result, Treasury faces a negative funding spread. The funding
spread varies over time and depends on Treasury bill rates relative to the
federal funds rate-increases in Treasury bill yields relative to the
federal funds rate increase Treasury's negative funding spread, and
declines in Treasury bill yields relative to the federal funds rate reduce
Treasury's negative funding spread.

As a result of the negative funding spread, Treasury strives to minimize
cash balances in order to reduce overall borrowing costs. To do this,
Treasury has worked toward improving cash forecasting. Treasury has
reduced the average cash forecasting error by one-half over the last 6
years according to Treasury officials. They credited improvements to
better technology and communication with their "lockbox banks" that
process certain tax payments. For example, the Electronic Federal Tax
Payment System provided actual cash flow information to Treasury
forecasters through electronic notification of pending tax payments,
replacing imperfect forecasts.

Increasing the earnings on excess cash balances is another way to narrow
the negative funding spread. Figure 9 below describes steps Treasury has
taken to increase earnings on its excess cash balances.

Figure 9: Steps Taken in Recent Years to Increase the Rate Paid on
Treasury's Cash Balances

aFederal Register, vol. 64, no. 146 (July 30, 1999), pp. 41748-41749.

bFederal Register, vol. 67, no. 51 (Mar. 15, 2002), pp. 11573-11577.

cKenneth D. Garbade, John C. Partlan, and Paul J. Santoro, "Recent
Innovations in Treasury Cash Management," Current Issues in Economics and
Finance, vol. 10, no. 11, Federal Reserve Bank of New York (November
2004).

CM Bills Are Generally Less Costly than Issuing Longer-term Regular
Securities

To make large payments when cash balances are low, Treasury can issue a CM
or a regular Treasury bill, although as explained earlier Treasury tries
to limit changes in the size of regular bill issuance from week to week.
When comparing the cost of a CM bill with the cost of a 4-week bill-the
regular bill with the shortest maturity-we see that the CM bill is
generally less costly for shorter-term needs (i.e., less than 28 days)
despite its higher yield. While the 4-week bill may have a lower yield,
the amount borrowed is generally outstanding for a longer period. For
example, Treasury issued a CM bill on June 3, 2005, for $16 billion. The
daily cost of borrowing was about $1.3 million. If Treasury had borrowed
$16 billion using the 4-week bill issued the day before, the daily
borrowing cost would have been only $1.2 million but the amount would have
been outstanding for 16 days longer and cost an additional $18.4 million
(see table 3). However, the extra cost of the 4-week bill would be
partially offset by the amount earned on cash balances.

Table 3: Illustrative Comparison of 4-week and CM Bill Borrowing Costs

Amount Term Yield Daily cost Total cost of
borrowed (days) of borrowing borrowinga
(billions of (percent) (millions of (millions of
dollars) dollars) dollars)
4-week bill $16 28 2.784 $1.2 $34.1
CM bill 16 12 2.979 1.3 15.7
Difference $18.4
(4-week bill -
CM bill)

Source: GAO analysis of BPD data.

aThis refers only to interest payments and excludes earnings on any excess
cash balances.

Note: The illustrative example shown above uses the actual yields paid on
the 4-week bill issued on June 2, 2005, and the CM bill issued on June 3,
2005. However, for comparability, we assumed bills of equal amounts (the
actual amount of the 4-week bill issued on June 2, 2005, was $20 billion).
In reality, the yield would be affected by the amount offered.

To compare the total cost of the 4-week bill with the cost of a CM bill
requires also looking at what happens to the cash that would have been
used to pay the maturing CM bill. CM bills are typically issued in months
with midmonth cash inflows. If Treasury issued a 4-week bill instead of a
CM bill in these months, these midmonth cash flows could be held in TT&L
accounts.48 However, since Treasury generally earns less on excess cash
balances than it pays to borrow, the additional borrowing costs associated
with the 4-week bill may not be completely offset.

Treasury Has Refined Its Use of CM Bills but Costs Have Begun to Increase

Treasury has taken steps to reduce use and overall cost of CM bills.
Lowering borrowing costs of CM bills could be achieved by combinations of
reducing the dollar amount issued or reducing the term to maturity (i.e.,
the number of days outstanding). Recognizing that CM bills were a
relatively costly way to absorb cash balance swings, Treasury introduced
the 4-week bill in 2001 to help reduce the use of CM bills. Treasury has
also increasingly issued CM bills in multiple tranches, which contributes
to smaller average issues, shorter terms to maturity, and lower total
borrowing costs. Borrowing costs declined in 2002 as Treasury reduced the
use of CM bills for longer-term borrowing (i.e., 28 days or more) and
short-term rates declined. However, borrowing costs associated with CM
bills have increased since 2003.

Treasury Initially Reduced the Amount of CM Bill Issuance Following the
Introduction of the 4-week Bill

Treasury has reduced reliance on CM bills since 2001. Initially, Treasury
reduced the total amount of CM bills issued by more than half from $346
billion in fiscal year 2001 to only $124 billion in fiscal year 2002 (see
fig. 10). This was the lowest amount issued in the previous 6 years. The
decline in CM bill issuance, however, was only temporary. Since 2002,
Treasury has increased its use of CM bills and in fiscal year 2005,
Treasury issued $268 billion in CM bills.

Figure 10: Amount of CM Bill Issuance (Fiscal Years 1996-2005)

In 2001, Treasury said that CM bills were not the most cost-efficient
means to absorb cash balance swings. At that time, the yield differential
was about 47 basis points and the amount of CM bills issued was about 20
percent of Treasury's short-term financing. Treasury introduced the 4-week
bill in 2001 to help reduce the need for CM bills by helping to smooth
swings in cash balances. In general, shorter-term securities provide
Treasury greater flexibility to adjust cash balances and outstanding debt
in response to actual cash needs. From fiscal year 2002 through 2005, CM
bills represented only 7 percent on average of Treasury's short-term debt
issuance. However, according to Treasury, the ability of the 4-week bill
to absorb cash balance swings is limited to swings in cash balances that
are longer than the two-week cash financing gap at the beginning of most
months.

Issuing CM Bills in Multiple Tranches Has Helped Reduce Borrowing Costs

Treasury has taken steps to borrow cash closer to the day it is needed,
which contributes to smaller average issues, shorter terms to maturity,
and lower borrowing costs. Since 2003, Treasury has increasingly issued CM
bills in multiple tranches-successive shorter-term CM bills that mature on
the same day. For example, instead of issuing one $30 billion or $40
billion CM bill on the 1st that matured on the 15th, Treasury might issue
three smaller CM bills on the 1st, 3rd, and 7th (all matured on the 15th).
As mentioned earlier, issuing in multiple tranches may increase the rate
paid, but it allows Treasury to borrow closer to the time when cash is
needed and ultimately reduces borrowing costs by reducing the average term
to maturity and annualized amount outstanding. Table 4 shows how issuing
in multiple tranches would have reduced borrowing costs in June 2005.

Table 4: Illustrative Effect of Issuing CM Bills in Multiple Tranches on
Borrowing Costs

Amount borrowed Term (days) Borrowing cost (millions of
dollars)
(billions of
dollars)
Multiple tranches
June 1, 2005 $10 14 $11
June 3, 2005 16 12 16
June 7, 2005 12 8 8
Total 38 34
Single tranche
June 1, 2005 38 14 43
Total difference -$8
(Multiple-Single)

Source: GAO analysis.

Note: The illustrative example shown above assumed Treasury would pay the
same discount rate on one CM bill as it did on the first part of the
multiple tranche. In reality, the discount rate would be different because
the rate depends on the amount auctioned, among other things. When
considering Treasury's total borrowing cost, the earnings on excess cash
balances held from the 1st-7th would partially offset borrowing costs.

The move toward multiple tranches has led to smaller CM bill issues on
average. Figure 11 shows that the average CM bill issue has declined by
more than half since 2001.

Figure 11: Average Size of CM Bills Issued (Fiscal Years 1996-2005)

Following introduction of the 4-week bill in 2001, Treasury reduced the
use of CM bills for longer-term borrowing (i.e., 28 days or more). Figure
12 shows that Treasury has reduced the average term to maturity (i.e.,
length of time outstanding) for CM bills issued by about half since 2001.
Prior to fiscal year 2002, Treasury issued CM bills with terms as long as
83 days. Since 2002, the longest maturing CM bill was 19 days.

Figure 12: Weighted Average Term to Maturity of CM Bills (Fiscal Years
1996-2005)

Note: Term to maturity of each CM bill is weighted by the amount issued.
Because CM bill issuance amounts vary, the weighted average maturity of CM
bills was calculated in order to reflect each CM bill's relative
importance in the total amount issued during the period.

Lower average issue sizes together with the reduced term to maturity
contributed to lower borrowing on an annual basis (see fig. 13). After
Treasury introduced the 4-week bill in 2001, the annualized amount of CM
bills outstanding declined dramatically from $19.2 billion in fiscal year
2001 to $3.3 billion in fiscal year 2002. However, since then the amount
has generally increased as Treasury increased the use of CM bills to
finance intramonth cash financing gaps. From fiscal year 2003 to fiscal
year 2004, the annual amount outstanding more than doubled to $8.1
billion. Treasury issued a similar amount of CM bills in fiscal year 2005.

Figure 13: Annualized Amount Outstandinga (Fiscal Years 1996-2005)

aThe annualized amount outstanding is derived by spreading each CM bill
across the 365-day year. So, for example, a 14-day CM bill with a face
value of $25 billion is multiplied by 0.038 (= 14/365) resulting in an
annualized amount outstanding of $959 million.

Borrowing Costs Declined Dramatically in 2002 but Have Since Increased

While Treasury's actions from fiscal years 2001 to 2002 reduced CM bill
borrowing costs, overall declining interest rates also helped. Borrowing
costs began declining in 2001 and dropped dramatically in 2002 (see fig.
14). Treasury's smaller CM bill issuances and reduced term to maturity
helped reduce the annual amount outstanding. According to our analysis,
these actions taken by Treasury contributed to about $610 million of the
reduction in borrowing costs from 2001 to 2002. We estimated that the
remaining decline was due to reductions in the yield paid on CM bills. As
noted earlier the yield differential varies with the overall level of
short-term yields in the economy. As these yields declined so did the
yield differential-by 43 basis points from fiscal year 2001 to 2002.

Figure 14: Total Borrowing Costs Associated with CM Bills (Fiscal Years
1996-2005)

However, since 2003 the borrowing costs associated with CM bills have
increased. Our analysis indicates that almost all of the increase in CM
bill borrowing costs is due to increasing rates. While Treasury cannot
control the overall level of short-term interest rates in the economy, it
can continue to take steps to reduce the use of CM bills and, according to
our analysis, the CM bill yield differential.

Options Exist That May Reduce the Use and Cost of CM Bills

We identified a range of options that may reduce the use or cost of CM
bills. The most promising options in our view-exploring ways to better
align cash flows and increasing the earnings on cash balances-are
discussed first. Other options, such as introducing a new shorter-term
regular bill and enhancing the transparency of CM bill auctions, are
discussed later.

Better Aligning Cash Flows Might Reduce Cash Financing Gap

Given that Treasury has increasingly used CM bills to fill regular cash
financing gaps, taking steps to better align large cash flows might help
reduce the use of CM bills. There are three ways to do this: smoothing the
payment of large federal expenditures, smoothing the payments of corporate
and nonwithheld individual tax payments, and aligning increased debt
issuance with large payments. If cash flows had been aligned to better
fill the frequent cash financing gaps, Treasury may not have needed the 11
CM bills it issued on the 1st through the 3rd of the month that matured on
the 15th of the month during fiscal year 2005 and could have reduced CM
bill borrowing cost by as much as $174 million or about 80 percent in
fiscal year 2005.49

Statutory and regulatory changes would be required to change the timing of
federal benefit payments and tax collections. While implementing any
changes is outside debt managers' control, Treasury could start a
discussion with other agencies and Congress to identify the costs and
benefits of alternatives to align or more evenly distribute federal
expenditures and tax receipts and seek any statutory authority necessary
to better smooth them. However, changing the timing of benefits and tax
collections, either jointly or independently, will have direct and
indirect effects not only on borrowing costs but also on individuals,
nonfederal entities, and federal government operations. These of course,
would have to be considered when making decisions about whether and how to
smooth federal cash flows. Looking at SSA's experience with spreading the
payments for new Social Security beneficiaries throughout the month may be
useful.

Treasury could also explore aligning increased borrowing capacity with
large payments in the beginning of the month to help reduce the cash
financing gap and reduce the use of some CM bills. Overall borrowing needs
are projected to be significant and grow over the long term and Treasury
will need to consider changes to its debt portfolio and auction schedule
to increase borrowing. Cash flow considerations can be a tiebreaker when
choosing between equally attractive alternatives. In the past, when making
changes to the range of maturities offered, Treasury has considered
aligning cash inflows with outflows in the beginning of the month. Going
forward, Treasury may reduce its reliance on CM bills-at least
temporarily-by continuing to consider adjustments that would better align
its increased debt issuance with its largest cash payments in the
beginning of the month.

Increasing Earnings on Cash Balances Would Reduce the Costs Associated
with Running Higher Cash Balances to Avoid Issuing CM Bills

Currently, Treasury faces a negative funding spread and as a result it is
more costly to maintain high cash balances to meet upcoming payments than
issue a CM bill. However, increasing the earnings on cash balances would
reduce the costs associated with running higher cash balances and may
ultimately reduce the use of some CM bills. Treasury previously explored
increasing the rate earned on TT&L accounts; however, banks objected.
Since then, Treasury introduced the TIO program,50 which pays higher rates
on cash balances than the TT&L program. Increasing balances in TIOs
relative to the TT&L accounts would help Treasury increase earnings on its
cash balances. However, Treasury would lose some flexibility because TIO
balances are not callable in the event cash balances unexpectedly fall
below cash needs.

Treasury could also explore broader options to increase earnings on cash
balances. For example, some countries' debt management offices engage in
reverse repurchase agreements. In a reverse repurchase agreement, Treasury
would lend market participants cash to purchase securities in the
secondary market. The borrower would then return the cash borrowed plus
interest at a specified time, usually overnight. Reverse repurchase
agreements could potentially narrow Treasury's negative funding spread
because repurchase agreement rates are generally higher than rates earned
on TT&L accounts. Treasury could explore the benefits and cost of
designing a new system to perform these transactions but implementing
reverse repurchase agreements would require legislative authority
according to Treasury.

Shorter-term Regular Bills Might Reduce the Need for Some CM Bills

Another way to smooth cash flows would be to introduce a shorter-term
instrument. The 4-week bill was partially successful in smoothing cash
flows throughout the year and reducing the use of CM bills. However, in
recent years, Treasury has increasingly used CM bills to fill cash
financing gaps that frequently occur in the first two weeks of the month.
Introducing a regular shorter-term bill might help fill the frequent cash
financing gap and reduce the use of CM bills further. A new bill that
matures on Thursdays like other regular bills would not give the Treasury
the same flexibility as CM bill issuance because the beginning-of-month
outflows and midmonth inflows often fall on different days of the week.
Alternatively, a new short-term security with specific issue and
redemption dates (rather than day of the week) might help the Treasury
manage cash flow gaps.

In Treasury's view, a shorter-term security would not likely generate
market interest in a way that would distinguish it, on a cost basis, from
CM bills. However, our analysis of CM and similar-maturing outstanding
bills in the secondary market shows that buyers pay higher prices (or
accept a lower yield) for short-term bills than Treasury currently accepts
(or pays) for CM bills. Market participants we spoke with expressed mixed
views on the demand for another short-term Treasury security. While some
said there may not be enough demand for a short-term instrument, others
said that bills maturing on the 15th, for example, would have natural
buyers and give investors more flexibility by offering another maturity
date for short-term instruments. Our analysis shows that CM bills maturing
on the 15th of months when tax payments are due likely have lower yields
than CM bills maturing on other days. Treasury can examine whether it
would obtain better prices on a shorter-term bill with issue and maturity
dates on specific days of the month.

Enhancing Transparency of CM Bill Auctions May Reduce CM Bill Yields

Increased transparency on the potential size of CM bills might improve
bidding in CM bill auctions and potentially reduce the yield Treasury pays
on CM bills. While market participants might expect Treasury to issue CM
bills in the beginning of the month, market participants told us that they
do not always know the size of the CM bill offering. As a result, they may
not bid as aggressively. However, there are tradeoffs to consider. In
order to provide market participants more advanced notice on the general
size and timing of CM bills, Treasury would have to improve its own cash
forecasting. Also, Treasury achieves lowest-cost financing in part by
providing the market with certainty. Trying to add certainty to CM bill
issuance would eliminate the benefit of flexibility they provide and may
actually increase borrowing costs.

Relaxing the 35-percent Cap on CM Bill Auction Awards May Reduce CM Bill
Yields

Treasury limits the maximum auction award to a single bidder to 35 percent
of the total amount offered to the public.51 The 35-percent cap is
intended, in part, to foster a liquid secondary market for a new issue by
ensuring adequate and wide distribution of the supply of a security among
investors and prevent temporary shortages or "short squeezes." However,
our analysis suggests that relaxing the 35-percent rule for CM bill
auctions (and thus allowing higher concentration) might promote more
aggressive bidding, improve auction prices for Treasury, and thus reduce
the borrowing costs associated with CM bills. There are a number of issues
to explore. For example, market participants may come to expect poorer
liquidity for CM bills, which may lead to less aggressive bidding over
time. Further, there would be a higher risk of a squeeze in the CM bills
market, although the risk would be relatively small in our view because CM
bill trading is sparse both before auctions and after auctions according
to our analysis. (See app. II for more information.) Treasury could
explore adjusting the per-bidder cap on an experimental basis and
determine whether there are benefits of relaxing the existing 35-percent
rule in CM bill auctions.

Exploring Other Countries' Short-term Cash and Debt Management Practices
May Provide Useful Insights

Lastly, exploring other countries' practices may provide useful insights.
For example, other countries' debt management offices use repurchase
agreements as a tool to support their cash management. Exploring other
countries' experiences may provide insights on the benefits and costs of
repurchase agreements for a central government. Although repurchase
agreements generally have slightly higher yields than CM bills, they
provide an alternative way to obtain cash for short periods, usually
overnight. In fiscal year 2005, Treasury announced that it was examining
the feasibility of a securities lending facility, which would operate much
like a repurchase agreement. Although this facility is still in its early
proposal process, Treasury generally intends to lend securities that are
in such short supply that they may threaten the settlement of Treasury
market transactions in a timely manner. In return, Treasury would receive
cash or bonds. While Treasury intends borrowers to use this facility at
their discretion and does not plan to use it for Treasury's own cash
needs, it might also consider how the facility would affect its cash
balances and whether the lending facility could be used to obtain cash for
very short periods.

Conclusion

In the face of persistent federal deficits and growing net interest costs,
reexamining debt management practices is warranted. Treasury has made
progress toward reducing the cost of CM bills, but it may be possible to
do more. This report presents options worth exploring that taken alone or
in combination may further reduce federal borrowing costs by reducing
either the use or the cost of unscheduled CM bills. CM bills will continue
to be a necessary debt management tool to meet unexpected cash needs when
Treasury has low cash balances or when Treasury is nearing the debt
ceiling. However, in recent years, Treasury has increasingly used CM bills
to fill cash financing gaps that frequently occur in the first two weeks
of the month. Our analysis indicates that the yield differential between
CM bills and outstanding bills of similar maturity has increased as
short-term rates have risen. If these rates rise further, as market
participants expect, and return to levels consistent with a longer-term
historical average, the CM bill yield differential is likely to rise above
levels seen in recent years. While Treasury does not vary its debt
management strategy in response to changing interest rates, it should be
mindful that increasing rates are likely to raise the relative cost of
unscheduled CM bills. As a result, Treasury should consider options,
including better aligning cash flows and increasing earnings on cash
balances, that may reduce the frequent use of CM bills and ultimately
overall borrowing costs.

Recommendation for Executive Action

We identified options that could potentially reduce the use and cost of CM
bills. We recommend that Treasury explore options such as those discussed
in this report and any others it identifies that may help Treasury meet
its objective of financing the government's borrowing needs at the lowest
cost over time. We recognize that there are a number of tradeoffs to
consider. In its exploration, Treasury should consider the costs and
benefits of each option and determine whether the benefits-in the form of
lower borrowing costs-to the federal government (and so to taxpayers)
outweigh any costs imposed on individuals, businesses, and other
nonfederal entities. Treasury should also consider how options may be
combined to produce more beneficial outcomes.

Implementing some of these options would require changes to statute or
regulations. If Treasury determines that any of these changes would be
beneficial, we encourage Treasury to begin discussions with relevant
federal agencies and the Congress about obtaining the necessary
authorities.

Agency Comments

We requested comments on a draft of this report from Treasury and the
Federal Reserve. In oral and written comments, Treasury generally agreed
with our findings, conclusions, and recommendations. Treasury said that it
is committed to continuing to explore ways to further reduce financing
costs through changes in the use of CM bills and that many of the options
we identified are embodied in its current debt management policy. Treasury
emphasized that statutory authority is needed for some options,
particularly changing the timing of receipts and expenditures and
improving earnings on excess cash balances. Treasury also suggested some
technical changes throughout the report that we have incorporated as
appropriate. Treasury's comments appear in appendix IV. In addition, the
Federal Reserve Board provided technical comments that we incorporated as
appropriate.

As you know, 31 U.S.C. S: 720 requires the head of a federal agency to
submit a written statement on actions taken to address our recommendations
to the Senate Committee on Homeland Security and Governmental Affairs and
to the House Committee on Government Reform not later than 60 days after
the date of this report. A written statement must also be submitted to the
House and Senate Committees on Appropriations with the agency's first
request for appropriations made more than 60 days after the date of this
report. Because agency personnel serve as the primary source of
information on the status of recommendations, we request that the agency
also provide it with a copy of your agency's statement of action to serve
as preliminary information on the status of open recommendations.

We are sending copies of this report to the Chairs and Ranking Members of
the House Committee on Ways and Means, the Senate Committee on Finance,
the House Committee on Financial Services, the Senate Committee on
Banking, Housing and Urban Affairs, and other interested parties. We will
also make copies available to others upon request. In addition, the report
will be available at no charge on GAO's Web site at http://www.gao.gov .
If you or your staff have any questions about this report please contact
Susan J. Irving at (202) 512-9142 or [email protected] or Thomas J. McCool
at (202) 512-2700 or [email protected] . Contact points for our Offices of
Congressional Relations and Public Affairs may be found on the last page
of this report. GAO staff making key contributions to this report are
listed in appendix V.

Sincerely yours,

Susan J. Irving Director, Federal Budget Analysis Strategic Issues

Thomas J. McCool Director, Center for Economics Applied Research and
Methods

Detailed Methodology and Findings of Statistical Analysis of Cash
Management Bill Yield Differential Appendix I

To identify which cash management bill (CM bill) features, if any,
explained the higher yields paid on CM bills, we performed a statistical
analysis of CM bills issued during fiscal years 1996-2005. The dependent
variable in the regressions was the difference between a CM bill's
investment yield at the time of auction and the investment yield on
similar-maturity Treasury bills (i.e, the "yield differential"), where the
latter is measured by the weighted average yield on Treasury securities
that mature immediately before and after the CM bill.1 We regressed the
yield differential on key CM bill features including announcement period,
term to maturity, issuance amount, and whether the CM bills were off-cycle
or a reopening of a previous issue. We also examined the effects of CM
bills with different issue and maturity dates, such as whether the CM
bills were issued on the 1st-3rd, matured on a large tax payment date, or
were issued during a debt issuance suspension period (DISP). While we
analyzed the yield differential for all CM bills issued during fiscal
years 1996-2005, we focused on the 55 CM bills issued during fiscal years
2002-2005 because the introduction of the 4-week bill in 2001 led to a
significant reduction in the amount and term to maturity of CM bills and
caused a structural change in the CM bill market.

Our empirical results suggest that several variables affected the yield
differential during the period studied. Lower yield differentials appeared
to be associated with lower short-term interest rates, relatively low CM
bill auction amounts, and maturity around large tax payment dates. There
is also some evidence that CM bills have higher yields when issued in
multiple tranches, which are successively issued CM bills with the same
maturity date. Despite these findings, the opportunity for Treasury to
achieve additional savings by further exploiting characteristics that
affect CM bill yields appears limited.

Previous Research Based on Earlier Data Suggests Some CM Bill Features May
Reduce Costs

The existing literature on CM bills and their costs is limited. In an
early study of CM bills issued from 1980 through 1988, Simon found that
from 1 day before to 1 day after announcement, the average interest rate
differential between CM bills and adjacent-maturity Treasury bills
increased by 20 basis points to a statistically significant 28 basis
points.2 An analysis of variance demonstrated that the increase in the
differential was greater for CM bills that had shorter maturities, were
part of larger issues, or had shorter when-issued periods.3 In a later
study, Simon found that from January 1985 through October 1991 CM bills
cost about 6 basis points more than regular bills.4 There have been many
fundamental changes in the way Treasury raises short-term cash since the
article's publication that may make its findings less applicable now.
These changes include the move from multiple-price to single-price
auctions, the growth of the repurchase agreement market and the prevalence
of off-cycle CM bills. Most important was the addition of the 4-week
Treasury bill to Treasury's regular borrowing schedule in 2001, which led
to a significant reduction in the amount and term to maturity of CM bills.

In a study of the effect of reopenings on the liquidity of Treasury bills,
Fleming included a binary variable identifying CM bill reopenings.5
Fleming studied Treasury bills issued from 1996 to 2000 and found that
reopenings of any kind had a positive and significant effect on yields.
Moreover, the CM bill variable was significantly positive for 13-week
bills, demonstrating that regular bills reopened as CM bills tend to have
higher yields. Fleming interpreted the results as showing that the
yield-reducing effect of enhanced liquidity is more than offset by the
yield-increasing effect of an increase in supply.

In a recent paper, Seligman evaluated the differential between CM bill
yields and the yields on Treasury bills with similar maturity dates.6
Seligman found that CM bills that were issued off-cycle or were large
(relative to outstanding Treasury bills of similar maturity) had higher
yields than other CM bills. In contrast, CM bills with longer durations or
that had 2-day notices before auction had lower yields than other CM
bills. This study suggested that Treasury could reduce the yields of CM
bills by avoiding off-cycle issuances, reducing their relative size,
issuing CM bills with longer terms, and giving 2 days notice in advance of
auctions.

However, the data in Seligman's study covered auctions held between 1990
and 1999. Seligman's sample did not cover more recent CM bill auctions
that occurred since 2001 when the 4-week bill was introduced and only very
short-term CM bills were issued. Hence, the findings of his study may not
apply to the current environment.

In another recent study based predominantly on auctions held before the
introduction of the 4-week Treasury bill, Christopher found that the cost
of CM bills, as measured by the spread between CM bill and repurchase
agreement yields, is negatively influenced by the time between the sale of
a security and the settlement date.7 The rationale offered for this
finding is that the delay allows administrative efficiencies. Christopher,
like Seligman, found that longer-maturity CM bills have lower yields than
shorter maturities. She suggests that the optimum maturity of a CM bill is
nearly 93 days. This finding highlights the inapplicability of earlier
research to the current environment when 4-week Treasury bills are
available to help meet short-term financing needs.8

Comparing CM Bill Yields with Secondary Market Yields on Similar-maturing
Treasury Bills

Choosing the appropriate reference point for CM bill yields is important,
and differed in earlier studies. Christopher focused on the difference
between CM bill and repurchase agreement yields instead of the difference
between yields on CM bills and Treasury bills with similar maturities,
which was the focus of Simon's earlier research, Seligman's, and our own
research. Although repurchase agreements and CM bills are both means to
obtain cash in the short-term, they have a major difference-CM bills, like
regular Treasury bills, are risk free whereas repurchase agreements issued
by private borrowers involve some risk. As a result, we focus on the
difference between CM bill and outstanding Treasury yields because it
provides a more direct indicator of the higher yield that Treasury pays
when issuing CM bills. Financial market analysts we spoke with agreed that
this yield differential measure was an appropriate focus for our research.

Specifically, our estimate of the yield differential is the difference
between a CM bill's yield and the average secondary-market yield on other
Treasury bills that are most similar (in terms of maturity) to the CM bill
on the day of auction. That is, we compare CM bill yields with yields on
the two nearest-maturing-one before and one after-Treasury bills.9 CM bill
yields were obtained from the Bureau of the Public Debt (BPD) while rates
on similar-maturity outstanding Treasury bills were obtained from the Wall
Street Journal (WSJ). For each Treasury bill, the bid and ask rates were
converted to yields and averaged. Next, the weighted average yield for the
two bills nearest in maturity to the CM bill was derived. The weights were
based on the relative difference in each bill's maturity date from that of
the CM bill, with the Treasury bill having a closer maturity date
receiving a greater weight and the weights summing to one. In the final
step, the weighted average Treasury bill yield was subtracted from the CM
bill auction yield to obtain the yield differential.

There are limitations of our yield differential estimate. For example, any
effect from the announcement of CM bills on yields for similar-maturing
bills is not captured. If the announcement of a CM bill increased the
yield on similar-maturing bills, then our estimate may be understated.
Also, in some cases, the surrounding Treasury bills we used could include
CM bills that were reopenings of regular Treasury bills. This would also
lead to an understatement of the yield differential because the yield on
the outstanding securities including CM bills would be higher than
outstanding securities that did not include CM bills. However, few CM
bills issued in fiscal years 2002-2005 were reopenings. Alternatively, if
the yield curve were upward sloping and concave, the curvature of the
yield curve would result in a positive estimate of the spread even before
considering the effects of other factors and create a positive bias in our
calculation of the yield differential.

The Introduction of the 4-week Treasury Bill Reduced CM Bill Maturities
and Their Relative Importance in Short-term Financing

Table 5 provides data on key attributes of CM bills, many of which we used
in an attempt to explain the yield differential. The table divides the
data into two subperiods. The first subperiod covers auctions held from
fiscal year 1996 through fiscal year 2001 before the introduction of the
4-week Treasury bill, while the second subperiod includes auctions held
from fiscal year 2002 through fiscal year 2005 after the 4-week bill's
introduction.

Table 5: Key Features of CM Bills before and after the 4-week Treasury
Bill's Introduction

Pre-4-week Post-4-week
Treasury Treasury
Bill Bill (Fiscal
(Fiscal years
years 2002-2005)
1996-2001)
Number of Number of CM
CM bill bill
auctions: auctions: 55
66
Mean Maximum Minimum Mean Maximum Minimum
CM bill yield 5.36 6.56 3.82 1.76 3.52 0.91
(percent)
Yield on 5.01 6.38 3.34 1.67 3.47 0.79
outstanding
bills with
similar
maturity
(percent)
CM bill yield 0.350 1.086 -0.038 0.091 0.553 -0.035
less yield on
outstanding
bills with
similar
maturity
(percent)
Percent 7.3 21.2 -0.7 6.0 35.3 -3.7
difference
between CM
bill yield and
yield on
outstanding
bills with
similar
maturity
Repurchase 5.39 6.57 3.99 1.76 3.53 0.94
agreement
yield
(percent)
CM bill term 27.4 83 1 9.6 19 2
to maturity
(days)
Amount of CM 21.6 42 6 14.8 32 4
bills issued
(dollars in
billions)
Ratio of bids 2.9 5.3 1.5 3.2 5.9 1.7
tendered to
the amount of
CM bills
issued (i.e.,
the
bid-to-cover
ratio)
Average amount 33.1 66.8 21.1 65.9 86.6 52.9
of outstanding
bills with
similar
maturity
(dollars in
billions)
CM bill issued 68.7 136.7 17.2 23.0 50.4 5.6
as a percent
of outstanding
bills with
similar
maturity
Number of days 2.4 9 1 1.5 4 0
from
announcement
to auction
Portion of CM 12.1 n.a. n.a. 20.0 n.a. n.a.
bills
auctioned
during DISPa
(percent)
Portion of CM 63.6 n.a. n.a. 49.1 n.a. n.a.
bills issued
on 1st-3rd
days of montha
(percent)
Portion of CM 21.2 n.a. n.a. 49.1 n.a. n.a.
bills maturing
on tax due
datea
(percent)
Portion of CM 18.2 n.a. n.a. 30.9 n.a. n.a.
bills issued
on 1st-3rd
days of month
that mature on
tax due datea
(percent)
Portion of CM 33.3 n.a. n.a. 20.0 n.a. n.a.
bills maturing
April 15-30a
(percent)
Portion of CM 18.2 n.a. n.a. 38.2 n.a. n.a.
bills maturing
on tax days
other than
April 15a
(percent)
Portion of CM 39.4 n.a. n.a. 85.5 n.a. n.a.
bills issued
off-cyclea
(percent)
Number of days 8.2 14 1 10.3 17 3
from CM bill
maturity to
end of month
Days from 1.6 5 0 1.3 3 1
auction to
settlement
Portion of CM 30.3 n.a. n.a. 49.1 n.a. n.a.
bills in a
multiple
tranchea
(percent)
Tail: high .022 .100 .000 .013 .040 .000
minus median
discount rate
(percent)
Portion of CM 12.1 n.a. n.a. 7.3 n.a. n.a.
bills
auctioned on
Wednesday and
maturing after
end of montha
(percent)

Source: GAO analysis of data from BPD and WSJ.

Note: n.a. = not applicable.

aBinary variable.

Table 5 highlights the significant difference in CM bill maturities
between the two subperiods. For the 66 CM bills auctioned in fiscal years
1996 through 2001, the average maturity exceeded 27 days while the
longest-maturity CM bill had a term of 83 days. In contrast, in the 55 CM
bill auctions held from fiscal year 2002 through 2005 the average maturity
of the CM bills was only 9.6 days while the longest-maturity CM bill had a
term of 19 days. In the more recent period even the maximum maturity of 19
days was much shorter than the average maturity of 27.4 days during the 6
years before 4-week Treasury bills were introduced.

Treasury also reduced its reliance on CM bills for short-term financing
after 2001. The average dollar amount of CM bills issued at each auction
declined from $21.6 billion in fiscal years 1996 through 2001 to only
$14.8 billion in fiscal years 2002 through 2005. In contrast, Treasury
increased issuance of regular short-term Treasury bills. Before fiscal
year 2002 the average amount of outstanding Treasury bills with similar
maturities to newly auctioned CM bills was $33.1 billion. The average rose
to $65.9 billion in the more recent period. As a result, newly auctioned
CM bills averaged less than one-fourth of the average amount of
outstanding Treasury bills with similar maturities during the more recent
subperiod compared with three-fourths during the earlier subperiod. This
reflects the importance of 4-week Treasury bills during the later period.

Many of the features listed in table 5 were represented by binary
variables set equal to 1 if a CM bill had the characteristic and 0 if it
did not. Binary variables include: auction during a DISP, issuance on the
1st-3rd days of the month, maturation on a tax due date and a combination
of issuance on the 1st-3rd days and maturation on a large tax due date.
Additional binary variables capture whether or not a CM bill matured
between April 15 and the end of the month and whether or not a CM bill was
part of a multiple tranche. For each binary variable, table 5 shows the
percentage of CM bills that had a particular feature. Among the most
notable change in these characteristics between the two subperiods was the
more than twofold increase in the share of CM bills issued off-cycle from
39 percent to 86 percent.

Our Study Focused on CM Bills Issued during Fiscal Years 2002-2005

Because of the dramatic decrease in CM bill maturities and reduced
reliance on CM bills for short-term financing after the introduction of
the 4-week bill, our effort to identify characteristics that might affect
the yield differential focused on CM bill auctions held from fiscal years
2002 through 2005. Column A of table 6 provides the estimated coefficients
and summary statistics for an equation that includes attributes that have
a significant effect on the yield differential. Using the same
specification for the 66 CM bill auctions for the earlier period that
extended from fiscal years 1996 through 2001 produced results that differ
significantly from the estimates for the more recent period. Testing this
specification for structural change using a Chow test resulted in an
F-statistic of 5.18, which permitted us to reject the hypothesis that the
relationship remained stable between the two periods at the .01 level of
confidence. This test result provided support for our decision to focus on
auctions held after the 4-week Treasury bill was introduced.

Table 6: Regressions Explaining the Yield Differential (Fiscal Years
2002-2005)

A B C D E
Variable Coefficient t-stat Coefficient t-stat Coefficient t-stat Coefficient t-stat Coefficient t-stat
C 0.001 0.04 -0.009 -0.24 -0.005 -0.13 0.019 0.37 -0.040 -1.21
CM bill issued 0.282a 3.30 0.253a 3.50 0.281a 3.31 0.268a 2.59 0.250b 3.14
as a percent of
the average
amount of
similar-maturity
Treasury bills
outstanding
CM bills -0.118a -3.99 -0.122a -3.72 -0.116a -4.20 -0.115a -3.54 -0.125a -4.29
maturing April
15-30
CM bills -0.117a -3.28 -0.119a -3.15 -0.115a -3.52 -0.113a -2.80 -0.125a -3.55
maturing on
large tax due
dates other than
April 15
Treasury yield 0.038b 2.28 0.038b 2.30 0.037b 2.12 0.038b 2.20 0.044b 3.05
Multiple-tranche 0.060c 1.91 0.055b 2.00 0.059c 1.92 0.058c 1.77 0.042 1.48
CM bills
Term to maturity 0.002 0.66
Days from 0.005 0.25
auction to
settlement
Off-cycle -0.018 -0.54
Number of days 0.034b 2.56
from
announcement to
auction
Included 55 55 55 55 55
observations:
R-squared 0.379 0.386 0.380 0.383 0.448
Adjusted 0.316 0.309 0.303 0.306 0.379
R-squared
S.E. of 0.081 0.082 0.082 0.082 0.078
regression
DW-Stat 1.793 1.794 1.810 1.792 1.808
Mean dependent 0.091 0.091 0.091 0.091 0.091
var.
S.D. dependent 0.099 0.099 0.099 0.099 0.099
var.

Source: GAO analysis.

Notes: Estimated using White heteroskedasticity-consistent standard errors
and covariance.

aSignificant at the 1 percent level. bSignificant at the 5 percent level.
cSignificant at the 10 percent level.

Yield Differential Is Positively Correlated with Level of Treasury Bill
Yields

Our analysis suggests that investors may require a proportionate rather
than an absolute differential as compensation for unscheduled CM bills.
While previous research has not examined whether the yield differential is
correlated with the overall level of Treasury bill yields, figure 8 in the
main text indicates that the differential may tend to move in the same
general direction as the level of secondary-market yields on Treasury
bills with similar maturity. The estimated coefficient of the Treasury
bill yield shown in column A of table 6 is 0.038, which suggests that a
1-percentage-point increase in the Treasury bill yield is associated with
a 3.8-basis-point increase in the yield differential. During fiscal year
2005, for example, the average yield on Treasury bills with maturities
comparable to newly auctioned CM bills was 2.42 percent and the yield
differential averaged 13.5 basis points. If the yields on
comparable-maturity Treasury bills had been 5 percent10 instead of 2.42,
the results imply that yield differentials would have been about 10 basis
points higher in fiscal year 2005 than they actually were. A continuation
in the rise in Treasury bill yields that began in 2004 is therefore likely
to result in an increase in the yield differential.

Yield Differential Is Positively Correlated with Auction Amounts

Our results show that as the supply of a CM bill rises relative to the
supply of similar investment alternatives, the relative price of the CM
bill declines and the yield differential increases. The results for the 55
auctions held from fiscal years 2002 through 2005 suggest that the yield
differential rises with an increase in the ratio of the amount of CM bills
auctioned to the average amount of similar-maturity Treasury bills
outstanding, as shown by the positive and significant coefficient of 0.282
for this variable. The variable's significant positive coefficient is
consistent with the results of Simon, Fleming, and Seligman. In studying
the relationship between auction size and yields, Seligman noted that an
increase in the relative amount of CM bills auctioned could have two
opposite effects on their relative yields. On the one hand, a higher
relative amount could increase liquidity and therefore reduce the yield
differential. On the other hand, a higher yield differential might be
necessary to attract sufficient investor interest when the amount of CM
bills being auctioned is large relative to outstanding Treasury bills with
similar maturities. Our results suggest the supply effect dominates.

The estimated coefficient-0.282-of the average amount of similar-maturity
Treasury bills outstanding implies that a $1 billion increase in the
amount of CM bills auctioned would raise the yield differential 0.43 basis
points in 2005, other things constant.11 This coefficient can also be used
to indicate how much higher the yield differential might have been if CM
bills had remained as important a source of short-term Treasury financing
in fiscal year 2005 as they were in the years before the 4-week Treasury
bill was introduced. Multiplying the coefficient of 0.282 by 0.687, which
was the average ratio of the amount of CM bills to the average amount of
similar-maturity Treasury bills from 1996 through 2001, rather than the
actual 2005 ratio of 0.190 raises the yield differential 14 basis points.
In other words, if CM bills were used as intensively in fiscal year 2005
as they were during earlier years before the 4-week Treasury bill's
introduction, the yield differential might have been 28 basis points in
2005 instead of the actual differential of 14 basis points.12

Maturity on or around Large Tax Payment Dates Reduces Yield Differential

Our work suggests that maturity on or around a tax due date, which is a
feature not studied in previous research, reduces the yield differential.
Table 6 shows that there is a 12 basis point reduction for CM bills that
mature from the 15th to the end of April. Similarly, CM bills that mature
on tax due dates other than April 15 (i.e., the 15th of March, June,
September, or December) tend to have lower yield differentials by nearly
12 basis points.13 This may be explained by empirical evidence suggesting
that regular Treasury bills whose maturity dates immediately precede
corporate tax payment dates have special value because corporate
treasurers may wish to invest excess cash in securities whose cash flows
can be used to liquidate cash liabilities.14

CM Bills Issued in Multiple Tranches May Require Higher Yield
Differentials

Financial market participants informed us that CM bills issued in several
tranches may lead to cautious bidding and therefore require higher yields
to attract investors. Multiple-tranche CM bills are CM bills that are
issued on different days within a short period that have the same maturity
date. Previous research has not studied this feature of CM bills. Reliance
on multiple-tranche CM bills increased from slightly less than one-third
of issues in the earlier subperiod to approximately one-half of CM bills
issued during fiscal years 2002-2005. To test whether multiple-tranche
issues affect CM bill yields, we included a variable identifying this type
of CM bills in the equation shown in table 6, column A.

The results suggest that CM bills that are part of a multiple tranche
might have yields that are 6.0 basis points higher than other CM bills.
However, while the coefficient is significant at the 10 percent level, it
is not significant at the 5 percent level.15 Moreover, the estimated
multiple-tranche coefficient is not as robust as other coefficient
estimates. For example, with the inclusion of a variable representing the
number of days advance notice before auction in column E, the coefficient
of multiple-tranche CM bills is no longer significant even at the 10
percent level. Accordingly, the evidence that multiple-tranche issues
increase the yield differential should be viewed with caution.

Yield Differential Is Not Affected by Timing of Issuance

We also tested whether the issuance of CM bills on the 1st-3rd days of the
month or during DISPs affected the yield differential. Because one-half of
all CM bills were issued on the 1st-3rd days of the month in recent years,
these issues might in some sense be considered more regular and thus
require lower yields than CM bills issued at other times. However, the
estimated coefficient of a variable identifying CM bills issued on the
1st-3rd days of the month was insignificant.16

CM bills have been a useful tool for Treasury when approaching the debt
ceiling. From fiscal years 2002 through 2005, 20 percent of CM bills were
issued during DISPs-a period for which the Secretary of the Treasury has
determined that obligations of the United States may not be issued without
exceeding the debt ceiling. While the coefficient of a variable
identifying CM bills issued during DISPs had a negative sign, it was not
significantly different from zero.17

Several Findings from Earlier Studies Are No Longer Supported

Studies by Seligman and Christopher identified several other variables
that had significant effects on the differential between yields on CM
bills and either the yield on regular bills with similar maturity or the
yield on repurchase agreements.18 As noted earlier, however, these studies
employed samples consisting either entirely or mainly of observations from
CM bill auctions held before the introduction of the Treasury bill in
fiscal year 2001. The only finding from earlier studies that appears to
remain valid in this new environment of shorter-maturity CM bills is that
an increase in the ratio of CM bills auctioned to the outstanding amount
of similar-maturing bills tends to increase the yield differential.

Both Seligman and Christopher found that CM bills with longer terms to
maturity were relatively less costly. The equation in Column B of table 6
includes a variable for a CM bill's term to maturity. In contrast to
earlier findings, we found that the coefficient for term to maturity is
positive and insignificant. This may be because CM bills are now
concentrated at the very short end of the maturity spectrum leaving little
room for increases in maturity to affect the yield differential.

Christopher found that the number of days between the auction of a CM bill
and the settlement date reduced the differential between the CM bill yield
and the yield on repurchase agreements. The explanation for this
relationship is that the delay allows administrative efficiencies.
However, the estimated coefficient of the number of days between auction
and settlement (in column C of table 6) is insignificant.

In the sample of CM bill auctions that Christopher studied, CM bills that
were auctioned on a Wednesday and had a maturity date beyond the end of
the month tended to have lower yields. Since the inception of the 4-week
Treasury bill and the truncation of CM bill maturities, only 4 of 55
issues were in this category. In an equation that also included the other
variables in column A of table 6, the coefficient for a binary variable
designating such issues had a t-statistic of only 0.161 which led to the
rejection of the hypothesis that CM bills auctioned on Wednesdays and
maturing after the end of the month have lower yields.

Instead of issuing a new security, Treasury may add to, or reopen, an
existing issue, increasing the amount outstanding of the issue. CM bill
reopenings are fungible with previously issued regular bills and may enjoy
their liquidity. Seligman found that CM bills issued off-cycle were
significantly more costly than those that reopened a previous issue.
However, contrary to Seligman's results, our analysis indicates that
off-cycle CM bills are not more costly than reopenings. Column D of table
6 shows the sign of the coefficient of the off-cycle variable is negative
rather than positive although, more relevantly, the coefficient is not
significantly different from zero. The descriptive statistics in table 5
show that 86 percent of CM bills issued during the more recent period were
issued off-cycle compared with about 40 percent in the earlier period.
Off-cycle issuance has become a regular feature of CM bills and does not
command an extra return.

Seligman also hypothesized that an increase in the number of business days
between CM bill announcement and auction would reduce the yield
differential. While his estimates suggest that 2 business days advanced
notice reduces the yield differential, he found that increasing the
announcement period beyond 2 days did not further reduce the differential.
In contrast, we found that the sign for the coefficient of the number of
days of advance notice was positive and significant, as shown in column E
of table 6. This has the unexpected implication that increasing the number
of business days notice before auction increases rather than reduces the
yield differential. Because markets usually penalize uncertainty, this
result appears counterintuitive and should be studied further to determine
whether it arose because of a statistical problem such as the correlation
between the advanced notice variable and an omitted variable that may be
unobservable.

Selected CM Bill Features Reduce Yields but Opportunities for Additional
Exploitation of These Features May Be Limited

Our analysis of CM bill yield differentials from fiscal year 2002 through
2005 revealed several features of CM bills that affect the yield
differential; however, Treasury's ability to achieve additional savings by
further exploiting these features may be limited. The average yield
differential was substantially lower during the period from fiscal years
2002 through 2005 than it was in the preceding several years. The two most
important factors behind the yield differential decline between the pre-
and post-2002 period were (1) the substantial decline in the general level
of short-term rates and (2) the major reduction in the ratio of CM bills
issued to the average amount of Treasury bills outstanding. The level of
short-term interest rates is largely determined by Federal Reserve policy
and market forces rather than Treasury. Treasury bill yields may continue
to rise and reach levels that prevailed in the early period, thereby
erasing the portion of the decline in the yield differential caused by
lower interest rates. However, because the 4-week bill is now a permanent
feature of Treasury's auction schedule and has reduced Treasury's reliance
on CM bills, the portion of the decline in the yield differential
attributable to the reduced ratio of CM bills to similar-maturity Treasury
bills is likely to endure. The experience from fiscal year 2002 through
2005 as a whole suggests that the yield differential could remain about 13
basis points below pre-2002 levels.

While Treasury may have some ability to change the relative amount of CM
and regular bills issued, the mix is affected by cash flow patterns
largely beyond its control. Aligning cash outflows and inflows could
reduce the amount of CM bills issued, but Treasury does not have authority
over the timing of all cash flows. However, Treasury does have control
over debt-related cash flows. Efforts to better align the timing of net
increases in debt with its largest cash payments could reduce both the
size of CM bills and the yield differential.

To the extent that Treasury can increase the share of CM bills that mature
on dates when taxes are due, additional savings might be achieved. Already
more than 60 percent of CM bills met this criterion in fiscal year 2005,
so the possibility of achieving further savings through this feature may
be minimal. Moreover, Treasury's short-term borrowing needs do not
uniformly occur shortly before large tax due dates.

Nearly one-half of CM bills issued since fiscal year 2002 have been issued
in multiple tranches. Our results provide somewhat limited evidence that
multiple tranche issues may be more costly. Even if such issues were to
require higher yields, however, multiple tranche issues have the advantage
of reducing borrowing costs by minimizing the duration of the amounts
borrowed in later tranches. For example, instead of borrowing $20 billion
for 10 days, Treasury might use two tranches to borrow $10 billion for 10
days and $10 billion for 9 days, thereby saving 1 day's interest on the
second tranche. Accordingly, the savings that multiple tranche CM bills
provide by reducing the number of days that interest is paid on later
tranches is likely to more than offset the higher yields such issues might
entail.

Finally, the positive correlation found between the number of days of
advanced notice and the yield differential appears counterintuitive and
probably should not be used as a basis for reducing the notification
period.

Analysis of CM Bill Auctions Appendix II

Given that CM bill auctions are innately less predictable than regular
Treasury bill auctions, it is not surprising that, by some measures, CM
bill auctions are less successful than regular auctions. Better auction
performance can be characterized by greater participation, a broader
distribution of awards, and more preauction activity in the when-issued
market. These factors theoretically improve Treasury's borrowing costs. By
most measures of participation and activity we examined, CM bill auctions
perform less well than 4-week bill auctions. However, some other measures
indicate that Treasury obtained a better price at CM bill auctions
compared with 4-week bill auctions and that there is stronger demand for
CM bills than 4-week bills. However, we did not find a statistically
significant relationship between any auction performance measures and
yield differentials.1 We did find that in more concentrated auctions,
Treasury was less likely to pay more than the yield indicated in the
when-issued market at the time of auction.

Trading Activity Was Lower Prior to CM Bill Auctions than 4-week Bill
Auctions

Treasury auctions are preceded by forward trading in markets known as
"when-issued" markets. The when-issued market is important because it
serves as a price discovery mechanism that potential competitive bidders
look to as they set their bids for an auction. When-issued trading reduces
uncertainty about bidding levels surrounding auctions and also enables
dealers to sell securities to their customers in advance of the auction so
they are better able to distribute the securities and bid more
aggressively, which results in lower costs to Treasury.

However, we found that trading activity is sparse before CM bill auctions.
We counted the number of preauction trades on the day of CM bill auctions
and found that when-issued trading is lower prior to CM bill auctions
compared with regular 4-week bill auctions (see table 7). There is
generally more activity after CM bill auctions than before the auction.

Table 7: Pre- and Postauction Trading Activity in CM Bills and 4-week
Bills (Fiscal Years 2002-2005)

Average number of
trades
CM bills 4-week
bills
Preauction Postauction Preauction Postauction
2002 250 283 419 315
2003 63 144 271 218
2004 94 154 226 220
2005 97 149 322 290
Average 2002-2005 112 173 309 261

Source: GAO analysis of GovPX data.

Note: GovPX data provide information on primary dealer transactions
through the interdealer brokers for all U.S. Treasury securities.

Yield Treasury Paid on CM Bills Generally Reflected Market Information at
the Time of the Auction

To evaluate the yield obtained at auction, Treasury's auction studies used
the difference between yields at auction and the contemporaneous
when-issued yield at time of an auction, usually 1:00 p.m. We refer to
this measure as the contemporaneous auction spread. According to Treasury,
this is a good benchmark to measure auction yields because potential
bidders have a choice between purchasing securities at auctions or
purchasing securities in the when-issued market. A negative spread (where
the auction yield is less than the contemporaneous when-issued yield)
indicates Treasury paid a lower yield than expected, which is a positive
auction result. Alternatively, a positive spread (the auction yield is
greater than the contemporaneous when-issued yield) indicates Treasury
paid a higher yield than expected from information in the market at the
time of auction. This would indicate a poor auction result. By this
measure, CM bill auctions performed better than 4-week bill auctions. The
contemporaneous auction spread of CM bill auctions from fiscal year 2002
through 2005 was approximately zero (see table 8), which implies that the
yield Treasury paid on CM bills generally reflected market information at
the time of the auction. In contrast, the auction spread for 4-week bill
auctions averaged a positive 0.3 basis points for fiscal years 2002
through 2005. The information derived from the auction spread depends in
part on

whether the when-issued market is liquid around the time of the auction.2
Because the level of trading activity is relatively sparse before CM bill
auctions, the information provided by the contemporaneous auction spread
may be limited.

Table 8: Average Spread between Auction and When-issued Yields at Time of
Auction (Fiscal Years 2002-2005)

Basis points
CM bills 4-week bills
2002 0.3 0.4
2003 -0.5 0.3
2004 -0.3 0.6
2005 0.4 0.1
Average 2002-2005 0.0 0.3

Source: GAO analysis of GovPX data.

CM Bill Auctions Are More Concentrated and Have Fewer Awarded Bidders than
4-week Bill Auctions

In order to lower borrowing costs, Treasury seeks to encourage more
participation in auctions. In general, large, well-attended auctions
improve competition and lead to lower borrowing costs for Treasury. For
example, in the 1990s Treasury switched its auctions from multiple-price
to uniform-price format in order to encourage more aggressive bids and a
broader distribution of auction awards. Treasury found that the share of
the auction awarded to the top five competitive bidders declined under the
new auction format to about 35 percent for 2-year notes and 36 percent for
5-year notes. Using similar estimates of bidder participation in CM bill
auctions, we found that the share of the auction awarded to the top five
bidders was 60 percent or higher in over half of CM bill auctions held
during fiscal years 2002-2005. The share exceeded 60 percent in only 18
percent (37 of 209) of 4-week bill auctions held during fiscal years
2002-2005.

Treasury limits the maximum auction award to a single bidder to 35 percent
of the offering in part to foster a liquid secondary market for a new
issue.3 A higher concentration could reduce competition and restrict a
security's supply in the secondary market preventing its efficient
allocation among investors and possibly generating a "short squeeze." The
term "short squeeze" is used by market participants to refer to a shortage
of a security relative to willing buyers for the same security. These
squeezes arise because Treasury allows dealers to sell the security short
to its customers (or other dealers) in the when-issued market before
securities are auctioned. Lack of competition could also result in lower
prices (higher yields) at auction. In contrast, we found generally
negative correlations between concentration measures and auction spreads
for CM bill auctions. In other words, in more concentrated auctions,
Treasury was less likely to pay more than the yield indicated in the
when-issued market at the time of auction. According to Treasury, a high
concentration ratio in CM bill auctions may imply some bidders really want
a particular bill, which drives the price up and yield down.

Not only are CM bill auctions more concentrated, but fewer bidders in
total are awarded CM bills than 4-week bills. According to Treasury
officials, short-term securities have a more limited audience. From fiscal
year 2002 through 2005, about 17 bidders on average were awarded CM bills
in each auction compared with 22 bidders in 4-week bill auctions. More
than half of the CM bill auctions had 16 or fewer awarded bidders. In
contrast, only about 4 percent of 4-week bill auctions had 16 or fewer
awarded bidders.

More Common Performance Measures Indicate High Demand for CM Bills but
Greater Uncertainty in CM Bill Auctions Compared with 4-week Bill Auctions

More commonly cited measures of auction performance, such as bid-to-cover
ratios and auction tails (difference between the high and average discount
rates) provide information on the demand for Treasury securities and the
dispersion of bids, but these measures are limited.4 The bid-to-cover
ratio is the ratio of the amount of bids received in a Treasury security
auction compared with the amount of accepted bids. In general, higher
ratios signal higher demand for the security being auctioned. From fiscal
year 2002 to 2005, the bid-to-cover ratio for CM bill auctions averaged
3.17 (see table 9). In contrast, the bid-to-cover ratio for 4-week bills
averaged only 2.30. This suggests a stronger demand for CM bills than
regular 4-week bills; however, these results should be interpreted with
caution. Some market participants suggested that a high bid-to-cover ratio
may arise because many dealers participate in CM bill auctions to fulfill
auction requirements that are less costly to meet by participating in
short-term CM bills auctions than in auctions of longer-term securities.

Table 9: Bid-to-cover Ratios for CM and 4-week Bills (Fiscal Years
2002-2005)

Average ratio of amount tendered to amount accepted
CM bills 4-week bills
2002 3.12 2.34
2003 3.45 2.32
2004 3.18 2.23
2005 3.02 2.32
Average 2002-2005 3.17 2.30

Source: GAO analysis of BPD data.

Auction tails-the number of basis points between the high and average
discount rates-are a measure of the dispersion of the bids. Auction theory
suggests that the more diverse the beliefs of the bidders and the more
uncertain they are about the demand for the bills, the more dispersed the
bids submitted. In contrast, narrower tails indicate strong bidding and
therefore lower costs to Treasury. When evaluating the auction tails of CM
bills compared with regular Treasury securities, we found that CM bill
tails were slightly larger than 4-week bill auctions (see table 10).

Table 10: Auction Tails for CM and 4-week Bill Auctions (Fiscal Years
2002-2005)

Basis points
CM bills 4-week bills
2002 1.71 1.66
2003 1.33 1.00
2004 1.00 0.80
2005 1.29 1.21
Average 2002-2005 1.27 1.17

Source: GAO analysis of BPD data.

Note: Auction tails are measured as the difference between high and
average discount rates.

In summary, most measures suggest CM bill auctions perform less well than
4-week bill auctions. However, the low participation and high
concentration of CM bill auctions do not explain why Treasury paid higher
yields on CM bills than investors paid for similar-maturing bills in the
secondary market.

Selected Bibliography Appendix III

Bikhchandani, Sushil, Patrik L. Edsparr, and Chif-fu Huang. "The Treasury
Bill Auction and the When-Issued Market: Some Evidence." Draft (Aug. 30,
2000).

Christopher, Jan E. "Determinants of the Spread between Repo and Cash
Management Bill Yields: Technical Report." Unpublished manuscript (2005).

Fleming, Michael J. "Are Larger Treasury Issues More Liquid? Evidence from
Bill Reopenings." Journal of Money, Credit, and Banking, vol. 34, no. 3
(August 2002): 707-35.

Garbade, Kenneth D. "Treasury Bills with Special Value." Fixed Income
Analytics (Cambridge, Mass.: MIT Press, 1996): 181-197.

Garbade, Kenneth D., and Jeffrey F. Ingber, "The Treasury Auction Process:
Objectives, Structure and Recent Adaptations." Current Issues in Economics
and Finance, vol. 11, no. 2, Federal Reserve Bank of New York (February
2005).

Garbade, Kenneth D., John C. Partlan, and Paul J. Santoro. "Recent
Innovations in Treasury Cash Management." Current Issues in Economic and
Finance, vol. 10, no. 11, Federal Reserve Bank of New York (November
2004).

Malvey, Paul F. and Christine M. Archibald. Uniform-Price Auctions: Update
of the Treasury Experiences. Department of the Treasury, Office of Market
Finance (Oct. 26, 1998).

Nyborg, Kjell G. and Suresh Sundaresan. "Discriminatory versus Uniform
Treasury Auctions: Evidence from When-issued Transactions." Journal of
Financial Economics, vol. 42 (1996): 63-104.

Seligman, Jason. "Does Urgency Affect Price at Market? An Analysis of U.S.
Treasury Short-Term Finance." Journal of Money, Credit, and Banking
(forthcoming).

Simon, David P. "Segmentation in the Treasury Bill Market: Evidence from
Cash Management Bills." Journal of Financial and Quantitative Analysis,
vol. 26, no. 1 (March 1991): 97-108.

Simon, David P. "Further Evidence on Segmentation in the Treasury Bill
Market." Journal of Banking and Finance, vol. 18 (1994): 139-51.

Comments from the Department of the Treasury Appendix IV

GAO Contacts and Staff Acknowledgments Appendix V

Susan J. Irving (202) 512-9142 Thomas J. McCool (202) 512-2700

In addition to the contacts named above, Jose Oyola (Assistant Director),
Richard Krashevski, Naved Qureshi, and Melissa Wolf made significant
contributions to this report. Jennifer Ashford also provided key
assistance.

(450390)

www.gao.gov/cgi-bin/getrpt? GAO-06-269 .

To view the full product, including the scope

and methodology, click on the link above.

For more information, contact Susan J. Irving at (202) 512-9142,
[email protected] or Thomas J. McCool at (202) 512-2700, [email protected].

Highlights of GAO-06-269 , a report to the Secretary of the Treasury

March 2006

DEBT MANAGEMENT

Treasury Has Refined Its Use of Cash Management Bills but Should Explore
Options That May Reduce Cost Further

One result of persistent fiscal imbalance is growing debt and net interest
costs. Net interest is currently the fastest-growing "program" in the
budget and if unchecked, threatens to crowd out spending for other
national priorities. This report was done under the Comptroller General's
authority and examines the Department of the Treasury's (Treasury) growing
use of unscheduled cash management bills (CM bills). GAO describes (1)
when Treasury uses CM bills and why, (2) the advantages and disadvantages
of CM bills, and (3) steps taken by Treasury to reduce the overall
borrowing costs associated with CM bills. GAO identifies possible options
Treasury could consider to reduce the use and cost of CM bills further.

What GAO Recommends

GAO recommends that Treasury explore options such as those identified in
this report and any others it identifies to help it achieve its
lowest-cost borrowing objective. If Treasury determines that any of these
changes would be beneficial, GAO encourages Treasury to begin discussions
with relevant federal agencies and the Congress about obtaining the
necessary authorities.

Treasury generally agreed with our findings, conclusions, and
recommendations and is committed to exploring ways to reduce financing
costs through changes in the use of CM bills.

Treasury makes large, regularly occurring payments, such as Social
Security and federal retirement payments, in the beginning of the month
and often receives large cash inflows in the middle of the month from
income tax payments and note issuances. Because regular bills alone are
not sufficient to fill these intramonth cash financing gaps, since 2002
Treasury has increasingly issued CM bills to bridge this gap. CM bills
allow Treasury to obtain cash outside of its regular borrowing schedule in
varying amounts and maturities, but Treasury pays a premium for doing so.
GAO's analysis found that Treasury paid a higher yield on CM bills than
that paid on outstanding bills of similar maturity in the secondary
market.

Treasury has taken steps to reduce the use and cost of CM bills. Treasury
added a 4-week bill to its regular auction schedule in 2001, which led to
reduced CM bill issuance, shorter terms to maturity, and lower borrowing
costs in 2002. Treasury has also fine-tuned CM bill issuance by borrowing
closer to the time when it needs cash. However, borrowing costs associated
with CM bills have increased since 2003.

Borrowing Costs Associated with CM Bills (Fiscal Years 1996-2005)

While Treasury has made progress towards reducing the cost of CM bills, it
may be possible to do more. GAO's analysis indicates that the yield
differential has increased as short-term rates have risen. If these rates
rise further, as market participants expect, so will the yield
differential. While Treasury does not vary its debt management strategy in
response to changing interest rates, it should be mindful of their effect
on the relative cost of unscheduled CM bills and explore options to reduce
the frequent use of CM bills and ultimately overall borrowing costs. GAO
identified options worth exploring such as any additional opportunities
for closer alignment of large cash flows; possible options for increasing
earnings on excess cash balances; and introduction of a shorter-term
regular instrument.
*** End of document. ***