Bond markets in Spring 2020 and the response of the...
Transcript of Bond markets in Spring 2020 and the response of the...
1
Bond markets in Spring 2020 and the response of the Federal Reserve
Annette Vissing-Jorgensen, University of California Berkeley and NBER
August 26, 2020
Abstract: This paper studies bond market dislocations during the initial phase of the COVID crisis in March and April of 2020. Distortions were large in Treasury and investment-grade corporate bond markets with yields increasing sharply (and much more than CDS) from March 9 through mid-March. Outflows from bond mutual funds peaked during the weeks of the largest price distortions and were likely contributors to downward price pressure on both Treasuries and corporates bonds. Funds facing outflows disproportionately sold Treasuries. Other large Treasury sellers included foreign central banks and hedge funds. I link the poor performance of Treasuries and investment-grade corporate bonds in mid-March to a disappearing safety-effect. Analysis of Federal Reserve interventions reveal that Treasury yields started falling with large daily Fed purchases of Treasuries in the days after March 18. Investment-grade corporate yields fell sharply after the Fed’s March 23 announcement, which included corporate bond purchases. In sharp contrast to the importance of large purchases in the Treasury market, corporate markets stabilized without immediate purchases and purchases made have been delayed and modest. I argue that providing liquidity requires large purchases (the Treasury market) but that one can stabilize a market with few purchases if the announcement itself moves fundamentals (perceived credit risk) enough to stop selling (the corporate market). I compare these lessons with how QE worked during the financial crisis.
1
Bond markets in Spring 2020 and the response of the Federal Reserve
Annette Vissing-Jorgensen, University of California Berkeley and NBER
December 8, 2020
(First version August 26, 2020)
2
1. Dislocations in Treasury and investment-grade corporate markets in March Yields increasing sharply, much more than CDS
2. Who were selling Treasuries? Why? • Bond mutual funds:
Massive outflows from bond mutual funds of all types (into money market funds). And funds disproportionately sold Treasuries
Can be rationalized by disappearing safety-effect • Foreign central banks • Hedge funds (domestic and foreign)
3. Fed facilities to stabilize bond markets and stimulate the economy • Treasuries: Purchase announcement did not stop yield increase, but very large
actual purchases did Providing liquidity requires large purchases Worked different than Treasury QE during financial crisis
• Corporate bonds: Purchase announcements lowered yields, actual purchases were delayed and modest Fund flows reversed, yields and CDS fell. ``Tail risk” promise sufficient to calm
markets down (along with fiscal policy and better virus news)
3
BOND MARKET DISLOCATIONS IN MARCH 2020
Treasury yields spiked in mid-March as S&P500 kept falling: 10-year yield +64 bps from 3/9 to 3/18
2000
2500
3000
3500
S&P5
00 in
dex
.51
1.5
210
-Yea
r Tre
asur
y C
onst
ant M
atur
ity R
ate
01jan2020 01mar2020 01may2020 01jul2020
10-Year Treasury Constant Maturity RateS&P500 index
4
Treasury yields spiked more at longer maturities
0.5
11.
52
2.5
01jan2020 01mar2020 01may2020 01jul2020
30yr 20yr 10yr 5yr 2yr 1yr
Nominal yields
5
Yield spike driven by higher real yields, not expected inflation or credit risk
10-year yield, adjusted for inflation and CDS: +103 bps from 3/9 to 3/18
-10
12
01jan2020 01mar2020 01may2020 01jul2020
Treasury yield Inflation SwapGovt CDS Treas yield-Infl swap-CDSTIPS
10-year
6
Investment grade corporate bond spread spiked, much more than its CDS
Difference increased more than 200 bps from 3/9 to 3/23! Peaks>300 bps on 3/23
01
23
4
01jan2020 01mar2020 01may2020 01jul2020
Yield spread (Inv Grade - 5-Year Treasury)CDS (Inv Grade, 5-year)
Investment Grade
7
High yield corporate bond spread follows its CDS more closely
24
68
1012
01jan2020 01mar2020 01may2020 01jul2020
Yield spread (High Yield - 5-Year Treasury)CDS (High yield, 5-year)
High Yield
8
Yield spread-CDS peaks at about the same value for HY bonds and IG bonds, despite much higher credit risk for HY
o Fall 2008: Yield spread-CDS much higher for HY than IG (Bai and Collin-Dufresne ’18)
See Haddad, Moreira and Muir (2020) for more on corporate market dislocations
01
23
4
01jan2020 01mar2020 01may2020 01jul2020
Yield spread-CDS, Inv GradeYield spread-CDS, High Yield
9
WHAT HAPPENED? WHO WERE SELLING? WHY?
I will show you data on selling
Other papers focus on who were not buying (enough): Dealers
• Duffie (2020): Proposes central clearing of Treasuries to overcome dealer balance sheet constraints
• He, Nagel and Song (2020): Model link between dealer balance sheet constraints and asset prices
10
TREASURY SELLING
Treasuries: US Financial Accounts, L.210/FU.210
Use flows tables of the US Financial Accounts to track ownership changes (as opposed to valuation changes)
Sellers: Rest of the world, mutual funds, households
11
Buyers: Fed, money market funds
12
-300
-200
-100
010
020
030
0Bi
llion
2000 2005 2010 2015 2020
Rest of the world
-300
-200
-100
010
020
030
0Bi
llion
2000 2005 2010 2015 2020
Mutual funds
-300
-200
-100
010
020
030
0Bi
llion
2000 2005 2010 2015 2020
Households, incl dom HFs
Quarterly net purchases of Treasuries (flows), 2000Q1-2020Q2
13
MUTUAL FUNDS: Drivers of Treasury selling (of notes/bonds)
• Large rotation from bond funds of all types to money market funds in March 2020
Source: My calculations based on ICI data.
14
• And when funds sold, they disproportionately sold Treasuries, ``horizontal slicing”
Taxable bonds funds sold 11% of their Treasuries but only 1% of their other bonds
15
March outflows from bond funds was an 8 sigma event:
Source: ICI data. Includes ETFs after 2013.
March 2020: -$265B-200
-100
010
0Bi
llion
2006 2008 2010 2012 2014 2016 2018 2020
Bond funds, Monthly flows, 2007M1-2020M7-2
00-1
000
100
Billio
n
2006 2008 2010 2012 2014 2016 2018 2020
Stock funds, Monthly flows, 2007M1-2020M7
16
Bond fund outflows peaked in same week as Treasury yields spiked:
Source: ICI data.
Week ending March 18 ==>
<== Week ending March 25
-125
-100
-75
-50
-25
025
Billio
n
January February March April May June July August
Bond funds, Weekly flows, 2020
17
Possible reason for bond fund outflows: Disappearing safety effect
Krishnamurthy and Vissing-Jorgensen (2011, 2012) safety effect (for Treasuries, but inherited to some extent by Aaa)
18
19
REST OF THE WORLD: Drivers of Treasury selling
1. Official sales: FX intervention 2. Private sales: Unwinding of hedge fund trades 3. Mix of official and private: Reduced currency hedging by investors who invest
abroad in USD, leading to private sales or FX intervention
We’re particularly interested in sales of notes/bonds since that’s where price distortions emerged (plenty of demand for bills from money market funds)
20
Split into official vs. private foreign sales
Source: US Financial Accounts L.210, F.210 (all foreign), BEA Table 9.1 (foreign official). Foreign private calculated as all foreign minus foreign official. Based on the last column, BEA assumes a 6.08% return on all Treasuries and 6.79% return on notes/bonds in 2020Q1.
• Foreign official sales of notes/bonds twice as large as private sales: FX intervention • Transactions data (TIC) gives a different picture:
Total foreign notes/bonds net purchases -$280B, -$82B official, -$198B private But holdings approach more accurate (Bertaut and Judson, 2014)
21
-300
-200
-100
010
020
030
0Bi
llion
2000 2005 2010 2015 2020
Rest of the world
-300
-200
-100
010
020
030
0Bi
llion
2000 2005 2010 2015 2020
Rest of the world, private
-300
-200
-100
010
020
030
0Bi
llion
2000 2005 2010 2015 2020
Rest of the world, official
Quarterly net purchases of Treasury Notes/Bonds (flows), 2000Q1-2020Q2
22
By country, top 20 sellers of notes/bonds TIC data. Duration of note/bond holdings is not available at the country level. I assume a 6.79% return on notes/bonds in column 4 (as used by the BEA in table above).
23
• FX intervention looks important for China, Brazil, Saudi Arabia, South Korea • Hedge funds likely driving sales from Cayman Islands, possibly Ireland, Bermuda,
Luxembourg SEC’s Form PF, hedge funds selling to US investors, % of Net Asset Value, 2019Q4:
But smaller role when calculating sales from holdings than from transactions: -$38B vs. -$138B for Cayman Islands: Someone else sold via Cayman
24
Several papers point to the unwinding of Treasury basis trades by hedge funds as a key factor behind Treasury market dislocations
• Schrimpf, Shin and Sushko (2020), Hauser (2020), Duffie (2020) • Barth and Kahn (OFR, 2020) cast some doubt on importance of this for pricing
The trade is as follows:
1. Enter short Treasury futures position to deliver Treasury, get cash at a future date 2. Buy Treasury security (the cheapest to deliver) 3. Fund the Treasury position using repo
Profitable if at initiation the CTD Treasury is cheap relative to the future and you manage to roll over the repo financing at a cheap rate and meet margin along the way
• In mid-March, margins increased and volatility increased, leading to trade unwinding
25
Reduction of $131B from March 3 to March 31, 2020
Down $131B March 3 to March 31
500
600
700
800
01jan2020 01mar2020 01may2020 01jul2020
Levered money short positions in Treasury futures
26
Schrimpf, Shin and Sushko (2020): • CTD Treasury was cheap relative to
futures and repo: Graph difference between the return on legs 1, 2 (implied repo rate, for new positions) and the cost of borrowing in leg 3 (actual repo rate)
Barth and Kahn (OFR, 2020): • CTD Treasury was expensive relative to
similar but non-deliverable Treasuries. • Perhaps the unwind was quite orderly?
27
Reduced currency hedging by investors who invest abroad, leading to private Treasury sales or FX intervention
Example: (Riksbank Financial Stability Report 2020:1, or BIS Bulletin No 1 2020) • A Swedish pension fund wants US investments but no currency risk • A Swedish bank borrows $ in short-term dollar market (e.g., via CDs or CP bought by a
prime fund in the US) • Fund and bank enter into FX swap: Fund gets $ (pay Kr) upfront, pays $ (get Kr) later. • Normally: Bank funding and FX swap rolled over • But if bank funding and FX swaps are not rolled over:
- Fund sells $ assets to close out the FX swap. Bank uses $ to pay off its $ funding. - Or fund buys $ for Kronor in the spot FX market, which may lead to FX
intervention with the central bank selling $ assets
Why not roll over?
1. Supply: Bank may have difficulty getting $ funding 2. Demand: Stock market crash Lower amount of US assets need hedging
28
FX swap pricing suggests clear dollar shortages
Source: Federal Reserve Financial Stability Report, May 2020
But: There was a dollar shortage in 2008 too and foreigners didn’t sell Treasuries back then
• Below we’ll see suggestive evidence that Fed’s USD swap lines did help lower Treasury yields, implying a role for the dollar shortage for Treasury selling during COVID
29
Danish pension & insurance sector reduced hedged USD exposure in March
Billion Kroner March 2020
Change in exposure Change in hedging
P&I sector bought $ for Kroner in the spot FX market • Downward pressure on Krone
exchange rate
• To support the Krone, Nationalbanken bought Kroner, selling EUR reserves worth 57B kroner in March
30
DID FED ACTIONS HELP IMPROVE MARKETS AND THE ECONOMY?
Possibilities:
1. Fed facilities helped (3/23, 4/9 announcements marked in graph below) 2. Cares Act fiscal stimulus mattered (3/24) 3. US growth rate of virus spread fell
2200
2400
2600
2800
3000
3200
S&P
500
0.1
.2.3
.4D
aily
gro
wth
rate
in n
ew C
OVI
D c
ases
01mar2020 01apr2020 01may2020 01jun2020 01jul2020
Daily growth rate in new COVID cases S&P 500
31
Fed actions, March-July 2020:
• Reductions in Fed funds target • USD swap facilities to provide dollars to foreigners • Facilities to stabilize money markets after outflows from prime funds • Programs to stabilize bond markets (Treasuries, MBS, corporate, munis, ABS)
March 3, 10 am:
• Fed funds target ↓ 50 bps to 1-1.25%
March 15, 5 pm:
• Fed funds target ↓ 100 bps to 0-0.25 pct • Primary credit rate ↓ 150 bps to 0.25 pct. Discount window borrowing encouraged. • Rate on dollar swap lines with BoC/BoE/BoJ/ECB/SNB ↓ 0.25 pct to OIS+0.25 pct.
84-day borrowing introduced. • $500B Treasury purchases, $200B MBS purchases
32
March 17, 10:45 am: Commercial Paper Funding Facility (CPFF) restarted
• Buying A1/P1 CP, 90-day, OIS+110 bps (and some A2/P2). $10B credit protection from Treasury.
March 17, 6 pm: Primary Dealer Credit Facility (PDCF) restarted
• Up to 90 day at primary credit rate
March 18, 11:30 pm: Money Market Mutual Fund Liquidity Facility (MMLF)
• Lends funds to banks to buy assets from prime money market funds • Up to 1-year at primary credit rate if backed by Treasuries/Agencies, otherwise add
100 bps. $10B credit protection from Treasury.
March 19, 9 am: Temporary dollar liquidity arrangements with other central banks
March 20, 10 am: Dollar swap lines with BoC etc.: Goes from weekly to daily operations
March 20, 11 am: MMLF expanded to munis
33
March 23, 8 am:
• Unlimited Treasury, MBS purchases. Agency CMBS now included in MBS purchases • $300B in lending, backed by $30B credit protection from Treasury, via:
1. Corporate bond purchases: Investment grade issuers only
Primary market (PMCCF): Interest rate “informed by market conditions” Secondary market (SMCCF): Pricing at “fair market value”
2. Term Asset-Backed Securities Loan Facility (TALF)
Fed lending against AAA-rated ABS backed by consumer/small business loans
3. CPFF, MMLF expanded with more muni debt. 4. Main Street Lending Program (MSLP) will be forthcoming
March 31, 8:30 am: Repo facility for foreign and international monetary authorities
• Objective to support Treasury (and other) markets. IOER+25 bps
April 6, 2 pm: Fed will provide term financing backed by PPP loans
34
April 9, 8:30 am:
a. Corporate bond purchases (plus TALF) expanded: Up to $850B, $85B credit protection. Fallen angels added.
b. Main Street Lending Program: Up to $600B, $75B credit protection. SOFR+250 to 400 bps.
c. Municipal Liquidity Facility (MLF): Up to $500B, $35B credit protection
April 27-July 23: Term sheets updated for MLF, MSLP, PPPLF, TALF, SMCCF, PMCCF
Identification:
• Timing of Fed announcements and purchases + cross-section of securities • Focus on 3/15, 3/23, 4/9 (no large effects of the others on bond markets)
35
FED IMPACT ON TREASURY MARKETS
• 3/15: Announcement fails to stop yields from increasing on 3/17, 3/18 • 3/23: Yield falls, some of this drop is causal based on intra-day data
But larger drop on 3/20. Why? Was policy not crucial for stabilizing markets? Yes!
2000
2500
3000
3500
S&P5
00 in
dex
.6.8
11.
21.
41.
610
-Yea
r Tre
asur
y C
onst
ant M
atur
ity R
ate
01feb2020 01mar2020 01apr2020 01may2020
10-Year Treasury Constant Maturity RateS&P500 index
Vertical lines mark 3/15, 3/23, 4/9
36
Intra-day Treasury returns, March 23, 2020
-.02
-.01
0.0
1.0
2
0 2 4 6 8 10 12 14 16 18 20 22 24Hour
10-Yr Treasury futures (TY)
-.03-
.02-
.01
0.0
1.0
2.0
3
0 2 4 6 8 10 12 14 16 18 20 22 24Hour
TLT ETF
37
Massive daily Fed purchases from March 19 helped bring Treasury yields down
Confounding factors? Corporate yields went up on March 20 and the stock market fell Unlikely that Treasury yield decline on March 19-20 due to stabilizing economic news
3456
7 1011121314 18192021
24252627
28 234
5
6
9
10111213
16
17
1819
20
23242526
27 3031123 6
789 1314151617 2021222324
27282930
3
45
67 10111213
14
1718
192021
24252627
28 2
34
5
69101112
131617
18
1920 23
24252627 30311
23
6
78
9
10
131415
1617
2021
22
2324
272829
30
020
4060
80D
aily
Fed
Tre
asur
y pu
rcha
ses,
$ B
0.5
11.
52
10-Y
ear T
reas
ury
Con
stan
t Mat
urity
Rat
e
01feb2020 01mar2020 01apr2020 01may2020
10-Year Treasury Constant Maturity RateDaily Fed Treasury purchases, $ B
38
Settlement of first USD Swap Lines on March 19, $162B, also lines up with turning point
• Reduced need to sell Treasuries to raise $ • We don’t know the factor mapping $ availabiliy to Treasury sales • March 19 settlement amount about the same as Fed purchases on March 19+20
3456
7 1011121314 18192021
24252627
28 234
5
6
9
10111213
16
17
1819
20
23242526
27 3031123 6
789 1314151617 2021222324
27282930
3456 910111213 161718
19
20 2324
25
26
27 30
31
1
2
3 6
78
9
10 131415
16
17 202122
23
24 27282930
-25
025
5075
100
125
150
175
200
Daily
cha
nge
in US
D liq
uidity
swap
out
stand
ing
0.5
11.
52
10-Y
ear T
reas
ury
Cons
tant
Mat
urity
Rat
e
01feb2020 01mar2020 01apr2020 01may2020
10-Year Treasury Constant Maturity RateDaily change in USD liquidity swap outstanding
39
Lessons about how Treasury QE worked during COVID crisis
• Treasury selling driven by liquidity needs, not loss of confidence in Treasuries o Bond fund outflows, disproportionate Treasury selling o FX intervention o Hedge funds unwinding levered trades o Perhaps reduced $ hedging by foreign investors
• March 15 announcement wasn’t enough to make others provide Treasury liquidity,
in expectation of selling to the Fed/getting USD via swap facilities o It took large actual purchases+large actual USD swaps to bring yields down o For “market functioning QE”, flow effects are crucial o Policies cannot be evaluated simply based on announcement effects on yields o If your announcement doesn’t stop the selling, you need to buy a lot!
40
• This is very different from how Treasury QE worked in 2008/2009 o Then we saw large announcement effect
Remember the 50 bps drop in the 10-year on March 18, 2009
o Treasury QE back then was not about providing Treasury market liquidity -- no
large mutual fund outflows, no large rest of the world Treasury selling o Different channels for affecting yields (signaling, increased Treasury scarcity
etc., see Krishnamurthy & Vissing-Jorgensen, 2011)
2.4
2.6
2.8
33.
210
-Yea
r Tre
asur
y C
onst
ant M
atur
ity R
ate
01feb2009 01mar2009 01apr2009 01may200
41
FED IMPACT ON MBS MARKETS: Worked more like Treasuries than corporate
MBS risk premium spikes on March 19 despite March 15 announcement
0.5
11.
5U
S M
BS F
ixed
Rat
e O
AS
01feb2020 01mar2020 01apr2020 01may202
Vertical lines mark 3/15, 3/23, 4/9
42
Massive daily Fed purchases from March 20 help lower MBS risk premium
• Large Fed purchases a day later than for Treasuries
And yields turn a day later than for Treasuries
34567 1011121314 1718192021 2425262728 23456
910
111213
1617
18
19
20
23
24252627
30
31123
678910
1314151617
2021222324 272829
30
34567 1011121314 1718192021 2425262728 23456 910111213
16
1718
19
20
23
24
25
26
27
3031
12
3 678
9
10
13
14151617
2021222324
27282930
010
2030
40Da
ily F
ed M
BS p
urch
ases
, $ B
0.5
11.
52
US M
BS F
ixed
Rat
e O
AS
01feb2020 01mar2020 01apr2020 01may2020
US MBS Fixed Rate OASDaily Fed MBS purchases, $ B
43
MBS sellers included highly levered REITS:
• Fed announcements on 3/15 apparently not enough to make them stop selling. • Faced MBS default risk due to COVID, MBS prepayments, repo funding problems
o Needed liquidity
MBS: US Financial Accounts, L.211/F.211
Sellers:
44
FED IMPACT ON CORPORATE MARKETS
3/23, 4/9 policy announcements involved corporate purchases and had large immediate effects on both IG and HY markets
• Causal effect based on intra-day data • 3/23: Larger effects on IG than HY. Announcement said IG purchases only
0.0
2.0
4.0
6.0
8.1
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Hour
LQD, cumulative return since announcement, March 23, 2020
0.0
2.0
4.0
6.0
8.1
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Hour
HYG, cumulative return since announcement, March 23, 2020
45
• 4/9: HY yields fall more on 4/9 than 3/23. 4/9 announcement added HY (and more IG)
0.0
2.0
4.0
6.0
8.1
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Hour
LQD, cumulative return since announcement, April 9, 20200
.02
.04
.06
.08
.1
5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20Hour
HYG, cumulative return since announcement, April 9, 2020
46
What about purchase effects for the corporate QE?
• Purchases started only on May 12 and have been small (total of $13B by end of Sept, all under the SMCCF)
• Very different picture than for Treasuries!
010
2030
40D
aily
Fed
cor
p bo
nd/E
TF p
urch
ases
, $ B
01
23
45
01feb2020 01mar2020 01apr2020 01may2020 01jun2020 01jul2020
Yield spread (Inv Grade - 5-Year Treasury) (Left axis)CDS (Inv Grade, 5-year) (Left axis)Daily Fed corp bond/ETF purchases, $ B
Yiel
d sp
read
/CD
S, P
erce
nt
47
Why were so small purchases needed?
• Fund flows reversed, yields and CDS fell. ``Tail risk” promise sufficient to calm markets down (along with fiscal policy and better virus news) Falato, Goldstein and Hortacsu (2020), for corporate bond funds: Corporate IG fund outflows slow after 3/23. HY fund outflows revert after 4/9
48
• Announcements improved fundamentals (credit risk) enough to stop the selling
Intra-day CDS data from Haddad, Moreira and Muir (2020):
49
• Very large general reduction in risk premia after March 23, 8 am announcement:
Intra-day, March 23, 2020
Knox and Vissing-Jorgensen (2020) maps this to a 2 pct point reduction in the 1-year equity premium on 3/23 (updating Ian Martin’s EP data and linking EP & VIXY)
50
• These effects could be due to any of the policies announced on March 23: Unlimited Treasury & MBS purchases plus corporate purchases But there was a role for the corporate purchases announced:
Gilchrist, Wei, Yue and Zakrajsek (2020) - Exploit bond eligibility rules in SMCCP:
Regression discontinuity around max 5 year remaining maturity
- 70 bps lower yields for eligible IG bonds within 14 days. - For the same company: 20 bps lower yields for eligible IG bonds
Lessons about how corporate bond QE worked during COVID crisis
• Announcements by themselves calmed markets. A lot like the ECB’s OMT • Announcements Improved fundamental enough to stop selling • Therefore, few purchases were needed
51
WHAT ABOUT REAL EFFECTS OF FED COVID FACILITIES?
• Large corporate issuance boom after corporate program announced, about $1T from March to June
• Real effects of corporate bond issuance?
o Darmouni and Siani (2020): Not for the large firms who issued bonds Funds raised used were to increase liquid assets and repay existing bank debt
1 23
45
67
8
9
10
11
12
1
2
3
4
5
6
7
8
010
020
030
0
2019 2019.5 2020 2020.5 2021
IG HY
US corp bond issuance, monthly
Source: SIFMA
52
• But from the ECBs Corporate Sector Purchase Program we learned that real effects may be indirect: SMEs benefitted o With corporate QE, large firms use up less of bank lending capacity (Grosse-
Rueschkamp, Steffen and Streitz (2019), Ertan, Kleymenova and Tuijn (2018))
• Bank lending during COVID: Increases for very large firms, not for SMEs
Chodorow-Reich et al (2020)):
53
• Would SMEs have done even worse without Fed programs? Large firms would have used credit lines more. Would that have mattered?
Test: Did SMEs that were customers of banks with higher (ex-ante) large firm credit line exposure borrow less during COVID than other SMEs?
Kapan and Minoiu (2020): Yes, for both lending standards, PPP loans and syndicated loans
54
Conclusions: • Fed programs were central for stabilizing markets • But moral hazard is now an even bigger concern