A Black Swan

in the Money Market

John B.
B Taylor
Stanford University

John C. Williams
Federal Reserve Bank of San Francisco

Bank of Canada Conference on Fixed Income Markets

September 12-13, 2008

The views expressed in this paper are solely those of the authors and should not
be interpreted as reflecting the views of the management of the Federal Reserve
Bank of San Francisco or the Board of Governors of the Federal Reserve System.
 Turmoil in Money Markets
Percent † On August 9, 2007,
money markets
6.0

50
5.0 lurched into turmoil,
turmoil
with overnight rates
swinging away from
4.0

Effective Federal Funds

3.0 T
Target t Federal
F d l Funds
3-Month LIBOR
F d
the Fed’s target rate
2.0
and longer-term
money market rates
1.0
Sep 06 Feb 07 Jul 07 Dec 07 May 08
rising sharply.
 A Black Swan in the Money Market
† In first half of 2007,
2007
Percent spreads on 3-month
1.2
inter-bank loans (relative
1.0 to OIS) averaged 8 bp.
with
i h a SD off 1 bp.
b
0.8
3-Month LIBOR OIS Spread

0.6
† Beginning on August 9,
0.4 2007 spreads
d shot
h t up.
0.2

0.0
† In the year since then,
the 3-month
3 month Libor-OIS
Libor OIS
-0.2
2002 2003 2004 2005 2006 2007 2008
spread has averaged
67 bp., with a SD of
p
17 bp.

Libor: London inter-bank offer rate.

OIS: Overnight indexed swap (proxy for average expected overnight rate)
Aim of Paper
† Analyze and measure the roles of
counterparty risk and liquidity risk in
term inter
inter-bank
bank lending rates during the
past year.

† Evaluate effects of Term Auction Facility

((TAF)) on term lending
g spreads.
p
Arbitrage-Free Pricing
‰ Absent risk
Ab i k and
d transaction
i costs, arbitrage
bi implies
i li that
h rates
on term inter-bank loans should equal the OIS rate.

† Example:
Bank A loans Bank B $1 million for one month.

Bank A funds this loan by borrowing $1 million each day from overnight fed
funds market.

Bank A hedges interest rate risk by entering in a overnight index swap,

agreeing to pay the counterparty the difference between the contracted
fixed rate and the average overnight fed funds rate over the next month.

† In the past, arbitrage has kept the spread between Libor and
OIS rate below 10 basis points.

† y, the spread
Today, p is 80 basis points.
p What aren’t banks taking
g
advantage of this opportunity?
 Counterparty or Liquidity Risk?
† Counterparty risk: late
or non-payment of
principal and/or interest.
Percent
2.0

1.6 Median (15 Banks)

3-Month LIBOR OIS Spread † Liquidity risk: funds may
be needed soon and
1.2
hard to obtain
elsewhere
elsewhere.
0.8

0.4
† Liquidity risk implies that
banks are passing up
0.0
otherwise profitable
Jan 07 Apr 07 Jul 07 Oct 07 Jan 08 Apr 08 Jul 08 opportunities to
“preserve balance
sheet.”
h t”
 CD-OIS Spreads Show Same
Pattern as Libor-OIS
† CDs are a major supply
Percent
of bank funding from
6.0 outside banking sector
5.6 and less affected by
5.2 li idi problems.
liquidity bl
4.8

4.4 † CDs, term federal funds,

4.0 3-Month CD and
d Eurodollars
E d ll show
h
same pattern as Libor.
3-Month LIBOR
3.6

3.2

28
2.8 † Libor has tended to be
2.4 below other term rates
Jan 07 Apr 07 Jul 07 Oct 07 Jan 08 Apr 08 Jul 08
since March 2008,
causingg some to q
question
the accuracy of Libor.
Money Market Turmoil in Europe
Percent
1.2

1.0

0.8
3-Month LIBOR OIS Spreads

0.6 US EU UK

0.4

0.2

0.0

-0.2
2004 2005 2006 2007

EU: Euro Libor and OIS; UK: Pound Sterling Libor and OIS.
Indicators of Counterparty Risk

† Credit Default Swap (CDS) rates

† Libor-Tibor spreads

† Libor
Libor-Repo
Repo spreads
Five-Year Credit Default Swaps
Major U.S. Banks
240
Bank of America
200 Citigroup
JP Morgan
Wells Fargo
160 Median

120

80

40

0
07:01 07:04 07:07 07:10 08:01 08:04 08:07

Strong co-movement in CDS rates across major commercial banks.

 Yen Libor vs. Tibor
Percent Percent Percent
1.2 1.2
LIBOR TIBOR Spread (Left Scale)
LIBOR OIS Spread (Right Scale)
3-Month TIBOR 3-Month LIBOR 1.0
10
1.0
0.8

0.8 .25 0.6

.20 0.4
0.6
.15 0.2

0.4 .10 0.0

.05
0.2
.00
00

0.0 -.05
1996 1998 2000 2002 2004 2006 2008 Jan 07 Apr 07 Jul 07 Oct 07 Jan 08 Apr 08 Jul 08

Tibor: Survey of Tokyo banks (4 of 16 in Libor survey).

Libor-Repo Spread as Credit Risk:
Unsecured vs. Secured Lending
Percent
2.0
3-Month LIBOR Repo Spread
3-Month LIBOR OIS Spread

16
1.6

1.2

0.8

0.4

0.0
Jan 07 Apr 07 Jul 07 Oct 07 Jan 08 Apr 08 Jul 08
 Liquidity Measures:
Term Auction Facility (TAF)
† Goal: restore functioning
Billions of $
of term inter-bank
Percent
250 lending market, in part
by reducing stigma
associated with discount
3-Month LIBOR OIS Spread 200
Fed TAF Balance
Total TAF Balance 150 window
i d b
borrowing.
i
100

50 † Begun in Dec. 2007,

1.2
0 expanded several times.
1.0
0.8
† 28-day collateralized
0.6 (discount window) loans.
0.4
0.2
† Rate set in single-price
0.0
Jan 07 Apr 07 Jul 07 Oct 07 Jan 08 Apr 08 Jul 08
auction (every 2 weeks).

† Synchronized with dollar

loans from ECB and SNB.
TAF Affects Composition,
Not Size of Fed’s Balance Sheet
Billions of $
900

800 Currency Swap

TAF
700

l
PDCF
Primary
Credit

600 l
Outright Holdings
of SOMA Repos

500

400
Jan 07 Mar 07 Jun 07 Aug 07 Nov 07 Jan 08 Apr 08 Jun 08
Econometric Evidence:
3-month Libor-OIS Spreads
† We examine effects of our three market
market-based
based measures
of counterparty risk and the TAF on bank term spreads.

† Theory is silent on timing of TAF effects on spreads, so we

consider alternative specifications.

† First specification:
p

Libor-OIS = c
+ a*RISK MEASURE
+ Σ5i=1 bi*TAF AUCTION DUMMY(t-i)
Econometric Evidence:
Libor-OIS Spreads
p
(similar results for CD & Term FF rates)

(1) (2) (3)

Median CDS 0.56

(0.07)
Libor-Tibor 4.58
(
(0.45)
)
Libor-Repo 0.70
(0.04)
TAF Auction -0
0.09
09 0.93
0 93 0.07
0 07
(sum of coefs) (0.27) (0.18) (0.15)
Adj. R2 0.52 0.59 0.85

Sample: 1/1/2007 – 8/8/2008. Newey-West HAC standard

errors in parentheses.
AR(1) Specification:
Libor-OIS Spread
p
(similar results for CD & term FF rates)

(1) (2) (3)

Median CDS 0.15

(0.08)
Libor-Tibor 0.53
(
(0.26)
)
Libor-Repo 0.08
(0.04)
TAF Auction -0
0.06
06 -0
0.08
08 -0
0.13
13
(sum of coefs) (0.05) (0.06) (0.05)
Adj. R2 0.98 0.99 0.98

Sample: 1/1/2007 – 8/8/2008. Newey-West HAC standard

errors in parentheses.
Econometric Evidence
† Based on three measures of term lending
spreads:

† Estimated effects of all three measures of

counterparty risk have the right sign and are in
most cases statistically significant.
significant

† Estimated effect of TAF ranges

g between
-29 basis points and +145 basis points;
negative estimated TAF effect is statistically
insignificant in only 1 case (-13 basis points).
Robustness Analysis:
Alternative Specifications

† Post Dec-11 TAF dummy variable (Wu)

† Include lagged lending spread and alternative

TAF dates (McAndrews, Sarkar, and Wang)
Alternative Specification (Wu 2008)
Post Dec-11 TAF Dummy Variable

† Test whether Libor-OIS spreads are lower

since announcement of TAF than before, after
controlling for CDS spread.
spread

† Assumes TAF permanently affects spread.

† Specification:
Libor-OIS
Libor OIS = c + a
a*CDS
CDS + b*TAF
b TAF_DUMMY
DUMMY

TAF_DUMMY = 1 after Dec. 11

OLS Regression with TAF Dummy:
Libor-OIS
bo O S Spreads
Sp ads
(similar results for other spreads)

(1) (2) (3)

Median CDS 0.58

(0.15)
Libor-Tibor 4.26
(
(0.41)
)
Libor-Repo 0.66
(0.04)
TAF Dummy -0
0.03
03 0.29
0 29 0.06
0 06
(0.11) (0.04) (0.04)
Adj. R2 0.52 0.74 0.85

Sample: 1/1/2007 – 8/8/2008. Newey-West HAC standard

errors in parentheses.
AR(1)
( ) Regression
g with TAF Dummy:
y
3-month Libor-OIS Spreads

(1) (2) (3)

Median CDS 0.15

(0.08)
Libor-Tibor 0.55
(0 26)
(0.26)
Libor-Repo 0.08
(0.04)
TAF Dummy
D -0.08
0 08 -0.08
0 08 -0.05
0 05
(0.01) (0.00) (0.02)
Adj. R2 0.98 0.99 0.98

Sample: 1/1/2007 – 8/8/2008. Newey-West HAC standard

errors in parentheses.
AR(1)
( ) Regression
g with TAF Dummy:
y
3-month CD-OIS Spreads

(1) (2) (3)

Median CDS 0.54

(0.15)
Libor-Tibor 1.21
(0 44)
(0.44)
Libor-Repo 0.16
(0.12)
TAF Dummy
D 0.04
0 04 0.14
0 14 0.14
0 14
(0.07) (0.15) (0.14)
Adj. R2 0.92 0.91 0.91

Sample: 1/1/2007 – 8/8/2008. Newey-West HAC standard

errors in parentheses.
Econometric Evidence:
Post Dec-11 TAF Dummy Variable
† Based on three measures of term lending
spreads:

† Estimated effects of all three measures of

counterparty risk have the right sign and are in
most cases statistically significant.
significant

† Estimated effect of TAF ranges

g from
-8 basis points to +44 basis points;
negative estimated TAF effect is statistically
significant in only 3 cases.
Alternative Specification based on
McAndrews-Sarkar-Wang (2008)

† Test whether Libor-OIS spreads change following TAF

“events” (announcements, auctions), after controlling
for contemporaneous change in CDS spread.

† Assumes TAF events have lasting effects on spreads

(through lags of spread in equation).

† Specification:
Libor OIS = c + a*Lag(Libor-OIS)
Libor-OIS a*Lag(Libor OIS)
+b*ΔCDS + d*TAF_EVENT_DUMMY
Results with Announcement Effects
and Lagged Spreads

Libor-OIS Term Fed CD-OIS

Funds-OIS
Change in Median CDS 0.18
0 18 0.12
0 12 0.43
0 43
(0.07) (0.08) (0.17)
TAF announcements -0.05 -0.02 0.02
(0 02)
(0.02) (0 02)
(0.02) (0 04)
(0.04)
TAF Operations -0.02 -0.02 -0.03
(0.01) (0.01) (0.03)
Adj. R2 0.98 0.98 0.92

Sample: 1/1/2007 – 8/8/2008. Newey-West HAC standard

errors in parentheses.
Results with Announcement Effects
and Lagged Spreads
† TAF announcements and d operations
i have
h
statistically significant effects on Libor-OIS
spreads
p in MSW specification.
p

† But, these findings are sensitive to choices of

lending spread and TAF operations dummy:
¾ Estimated effects of TAF announcements is
insignificant using Term Fed Funds and CD
spreads.
¾ Estimated effect of TAF operations is insignificant if
TAF settlement dates are included in TAF operations
dummy.
Reconciling Results
† Th
The evidence
id off significant
i ifi t effects
ff t off TAF announcements
t
and operations on term lending spreads based on
specification with lagged spread appears to contradict
evidence from specification with post-Dec.
post Dec 11 TAF
dummy, which indicates that spreads are NOT much
lower after the introduction of the TAF.

† Evidently, on days without TAF announcements or

operations, spreads tend to rise, offsetting beneficial
effects of TAF announcements and operations.
operations

† These results are consistent with our first model, which

i
implies
li ththatt TAF effects
ff t on spreads
d are short-lived.
h t li d
Conclusion

† Risk measures are economically and

statistically significant predictors of term
lendingg spreads.
p This is a robust finding.
g

† We do not find similarly robust evidence of an

economically and statistically significant effect
of the TAF on spreads.

† Counterparty risk appears to be the

predominant source of the extraordinary
sustained rise in term lending spreads over the
past y
p year.