July 2009
Regulating Systemic Risk
(Abridged)
Viral Acharya, Lasse Heje Pedersen, Thomas Philippon, and Matthew Richardson
AQR Capital Management, LLC |Two Greenwich Plaza, Third Floor | Greenwich, CT 06830 |T: 203.742.3600 | F: 203.742.3100 |www.aqr.com
New York University Stern School of Business
Policy Proposal Part of NYU Stern Project and Follow-Up Research (Work in Progress)
Based on Chapter 13: “Regulating Systemic Risk”
2
Viral Acharya, Lasse Heje Pedersen, Thomas Philippon, and Matt Richardson
http://whitepapers.stern.nyu.edu/
Summary of our key findings
It is time to quantify systemic risk of financial institutions…
We propose a systemic risk measure: Marginal Expected Shortfall (MES)
• Average loss suffered by an institution when the market is in its left tail (say 5%)
Key findings based on MES:
• Securities dealers and brokers have been the most “systemic” institutions, every year, for past 45 years!
• MES of financials estimated during June 06-07 predicts their performance during ongoing crisisg p p g g g
• Top 4 out of 10 firms ranked by MES pre-crisis have effectively failed (Bear, Lehman, ML, CIT), tworeceived government support (GS, MS), others interesting (eTrade, CBRE, Charles Schwab, Ameritrade)
Systemic risk measures such as MES can be used to gauge which institutions are likely to suffer during Systemic risk measures such as MES can be used to gauge which institutions are likely to suffer duringaggregate crises (financial and/or economic) and potentially affect others
Systemic risk externalities can be regulated by “tax”ing institutions for their systemic risk contributions
We provide a scheme to calculate such systemic risk “tax” or the cost of insuring systemic risk of financials We provide a scheme to calculate such systemic risk tax or the cost of insuring systemic risk of financials
• Insurance cost based on June 2007 ranks at top Bear, Lehman, ML, MS, Fannie, Freddie, GS, Citigroupand JP Morgan
3
Systemic risk (MES5) over time for different groups
Annual Value-Weighted Marginal ES(5%) by groups.
0 11
0.09
0.11
0.05
0.07
0.01
0.03
-0.011963 1968 1973 1978 1983 1988 1993 1998 2003 2008
NBER Recession Depositories Others Insurance Sec. and Comm.
Ex Ante Systemic Risk (MES5) and Performance During Crisis
UBHCB K
.5c0
8
PB CT
BE RBRK
CB SSW FC
A TAGE
AFLCB
SA F
TRV
AOC
B RKCB H
B LKCG
B OT
M A
0ul
y07
to D
ec
TROW
BRK
S NVLUK
C IN F
LTRM TB
W FCM MC
JPMHUM
U NP
BK
N TRSTRV
PNCTM K PGR
USB
SEIC
UNMBBT
STT
SCHW
A LL
NYBA MTD
B LK
M ETAET
FIS
W LP AIZ
FNF
IC E
M A
W U N MX
U NH5ng c
risis
: Ju
C MA
E V
FITBRF
B EN
HB ANCNALNC
M IAX PBACC I
K EYLM
S TI
SEIC
M SC VHHNT
C OF
HIG
ZIONGS
JN S
W LP
P FGPRUCM E
AIZ
AM P NYX
U NH-.5R
etur
n du
ri
SOVW B
FNM
M ER
NCCCFCSLM
A IGBS C
CMB I
FREA BK LE HW M
HIG
E TFCA CAS CITGNW CBG
-1
0 .01 .02 .03 .04
5
MES5 measured June06 to Ju ne07
Ranking by systemic risk (MES) and cost of systemic insurance
MES (per share & $) Systemic Insurance Fee (% of equity & $) Ranking as of June 07 MES (%) MES ($) Ranking as of June 07 Fee (% Fee ($)E TRADE FINANCIAL CORP 1 37 BEAR STEARNS COMPANIES INC 1 6BEAR STEARNS COMPANIES 2 20 LEHMAN BROTHERS HOLDINGS INC 2 3LEHMAN BROTHERS HOLDINGS INC 2 3C B RICHARD ELLIS GROUP IN 3 54 MERRILL LYNCH & CO INC 3 2LEHMAN BROTHERS HOLDING 4 12 MORGAN STANLEY DEAN WITTER & 4 1MORGAN STANLEY DEAN WIT 5 4 COUNTRYWIDE FINANCIAL CORP 5 8GOLDMAN SACHS GROUP INC 6 5 FEDERAL HOME LOAN MORTGAGE C 6 4MERRILL LYNCH & CO INC 7 6 FEDERAL NATIONAL MORTGAGE AS 7 7SCHWAB CHARLES CORP NEW 8 16 GOLDMAN SACHS GROUP INC 8 5C I T GROUP INC NEW 9 50 E TRADE FINANCIAL CORP 9 15T D AMERITRADE HOLDING CO 10 42 C I T GROUP INC NEW 10 17T ROWE PRICE GROUP INC 11 36 AMERIPRISE FINANCIAL INC 11 14AMERIPRISE FINANCIAL INC 11 14EDWARDS A G INC 12 68 S L M CORP 12 20FEDERAL NATIONAL MORTGA 13 8 COMMERCE BANCORP INC NJ 13 21JANUS CAP GROUP INC 14 76 HARTFORD FINANCIAL SVCS GROUP 14 12FRANKLIN RESOURCES INC 15 13 METLIFE INC 15 11LEGG MASON INC 16 44 SOVEREIGN BANCORP INC 16 22AMERICAN CAPITAL STRATEG 17 62 UNUM GROUP 17 23STATE STREET CORP 18 24 WASHINGTON MUTUAL INC 18 16COUNTRYWIDE FINANCIAL CO 19 27 PRUDENTIAL FINANCIAL INC 19 13EATON VANCE CORP 20 75 JPMORGAN CHASE & CO 20 9
6
O C CO 0 5 JPMORGAN CHASE & CO 20 9
July 2009
Regulating Systemic Risk
(Unabridged)
Viral Acharya, Lasse Heje Pedersen, Thomas Philippon, and Matt Richardson
New York University Stern School of Business
AQR Capital Management, LLC |Two Greenwich Plaza, Third Floor | Greenwich, CT 06830 |T: 203.742.3600 | F: 203.742.3100 |www.aqr.com
New o U ve s ty Ste Sc oo o us ess
Overview: What to Do About Systemic Risk
Systemic risk: definition• The joint failure of a significant part of the financial institutions• Leading to the freezing of parts of the capital markets• That has the potential to disrupt the real economy
Systemic risk is very damaging• Losses of about 15 20% of GDP during banking crises over past 25 years• Losses of about 15-20% of GDP during banking crises over past 25 years.– Hoggarth, Ricardo and Saporta, “Costs of Banking System Instability: Some Empirical Evidence”, Journal of
Banking and Finance (2002)
• Systemic risk is different from risk: Lehman 08 vs. Barings 95
Key drivers:• Inability of private sector to resolve individual bank failures• Financial sector’s central role in the economy• Bailouts : Time-inconsistency problem
Moral hazard
• “Too big to fail” size bias; “Too interconnected to fail” counterparty risk bias;
8
• Too-big-to-fail size bias; Too-interconnected-to-fail counterparty risk bias;
“Too-many-to-fail” herding, systemic risk bias
Why Regulate Systemic Risk : Externalities
Spillover to the real economy, credit unavailability, payment system, etc.
• E.g., Acharya (2001, 2009), Diamond and Rajan (2005, 2009)
Market and funding liquidity spirals Market and funding liquidity spirals
• E.g., Geanakoplos (1997+), Brunnermeier and Pedersen (2009)
Fire sales and depressed prices can lead to allocation inefficiencies
• E.g., Acharya and Yorulmazer (2008)
What to do about systemic risk: treat it like pollution• Private regulation of systemic risk not feasible
St 0 A i t “ t i i k l t ”• Step 0: Appoint a “systemic risk regulator”– Central Bank a natural candidate (but issues of ensuring Central Bank independence must be addressed)– Macro-prudential regulation to supplement micro-prudential supervision by functional regulators
• Step 1: Measure systemic risk• Step 2: Require insurance against systemic risk, effectively taxing it and limiting it
Costs of not containing systemic risk: Without regulation there is
• Excessive leverage loading on aggregate risk concentration in illiquid assets
9
• Excessive leverage, loading on aggregate risk, concentration in illiquid assets
Measuring Systemic Risk
Key feature:
• For each bank: measure its contribution to a general crisis
Standard risk management calculation:
• Marginal Expected Shortfall (MES)
• Take 5% worst-case periods for the system (output, stocks, credit, financial sector,...)
• Ask: in those periods, how much did firm j contribute?
Analogy
• Allocation of economic risk capital within a firm– Each desk is charged for its (implicit) use of the firm’s economic capital
• Allocation of capital requirements within an economy– Government capital is a public goodp p g
Other measures:• ES: Expected shortfall of a firm based on own distribution
B t
10
• Beta
• Adrian and Brunnermeier (2008)’s CoVaR measure
A “demo” for the ongoing crisis
US financial institutions
• SIC code 6+
• Market capitalization > $5bln as of July 2007 (100 firms in total including all the usual suspects)Market capitalization > $5bln as of July 2007 (100 firms in total, including all the usual suspects)
Value-weighted CRSP Market returns
• MES based on 5%ile and 10%ile of Market returns
• Robust to employing financial sector as the “Market”
Pre-event period for systemic risk measurement: June 06 – June 07
Event period to explain realized performance: July 07 – Dec 08
Analysis of systemic risk measures (I-III for ongoing crisis; IV-VI for historical data)
I. Summary of measures and correlation
II. Predictive power of measures for realized systemic risk
III Cross sectional variation across types of institutions and bank characteristicsIII. Cross-sectional variation across types of institutions and bank characteristics
IV. Cyclical properties and stability of measures over time
V. Predictive power for crises
VI. Evidence on past crises
11
p
VII. Pricing of insurance against systemic risk contribution of individual institutions
I. Summary of measures and correlation (Table 1)
Panel APanel ADescriptive statistics of the measures Event return, beta, ES, MES5 and MES10.
Event
Return ES Beta MES5 MES10
Average -47% 2.73% 1.00 1.63% 1.26% Median -46% 2.52% 0.89 1.47% 1.17%Std. dev. 34% 0.92% 0.37 0.62% 0.48%
Min -100% 1.27% 0.34 0.39% 0.31% Max 36% 5.82% 2.10 3.36% 2.51%
Panel BPanel B
Sample correlation matrix of the measures Event return, beta, ES, MES5 and MES10
Event Return 1.00 ES -0.18 1.00
Beta -0.27 0.77 1.00 MES5 0 31 0 71 0 92 1 00MES5 -0.31 0.71 0.92 1.00MES10 -0.29 0.64 0.92 0.92 1.00
12
II. Predictive power for realized returns during the ongoing crisis
What works:
• Systemic risk measures– MES5
– MES10
– Beta
Wh d k ll What does not work as well:
• Firm-level risk measure– ES
In a horse race:
• MES5 and MES10 contain more relevant information than Beta
With what lead?
• Measures have predictive power up to 3-4 months lead time
13
Ex Ante Systemic Risk (MES5) and Performance During Crisis
UBHCB K
.5c0
8
Y = - 0.2 – 16.78 * MES5 + error, adj R2=8.70% (-2.24) (-3.26)
PB CT
BE RBRK
CB SSW FC
A TAGE
AFLCB
SA F
TRV
AOC
B RKCB H
B LKCG
B OT
M A
0ul
y07
to D
ec
TROW
BRK
S NVLUK
C IN F
LTRM TB
W FCM MC
JPMHUM
U NP
BK
N TRSTRV
PNCTM K PGR
USB
SEIC
UNMBBT
STT
SCHW
A LL
NYBA MTD
B LK
M ETAET
FIS
W LP AIZ
FNF
IC E
M A
W U N MX
U NH5ng c
risis
: Ju
C MA
E V
FITBRF
B EN
HB ANCNALNC
M IAX PBACC I
K EYLM
S TI
SEIC
M SC VHHNT
C OF
HIG
ZIONGS
JN S
W LP
P FGPRUCM E
AIZ
AM P NYX
U NH-.5R
etur
n du
ri
SOVW B
FNM
M ER
NCCCFCSLM
A IGBS C
CMB I
FREA BK LE HW M
HIG
E TFCA CAS CITGNW CBG
-1
0 .01 .02 .03 .04
14
MES5 measured June06 to Ju ne07
Ex Ante Systemic Risk (Beta) and Performance During Crisis
UBHCBK
.5c0
8
Y = -0.23 – 0.24 * beta + error, adj R2=6.19% (-2.45) (-2.77)
PBCT
B ERBRK
CBSSW FC
AT
UNP
AGEAFL
CB
SA F
TRV
AOC
BRKCBHBLK
CG
BOT
MA
0ul
y07
to D
ec
TROWSNV
LUK
CINF
LTRMTB
MMC
JP MHU M
UNP
B K
NTRSTRV
P NCTMKPGR
USB
SEIC
UNMB BT
STT
SC HW
A LL
NYBAM TD
BLK
META ET
FIS
W LP A IZ
FNF
ICE
MA
W UNM X
U NH5ing
cris
is: J
u
CMA
EV
FITBRF
B EN
HBA NCNALNC
MIA XPBAC CI
KEYLM
SLM
STI
MSCVHHNT
COF
HIG
ZIONGS
JNS
W LP
PFGPRU
CM E
A IZ
AMP NYX
U NH-.5R
etur
n du
r
SOV W B
FN M
MER
NCCCFCSLM
A IGBSC
CMBI
FREAB K LEHW M
HIG
ETFCACAS CITGNW C BG
-1
.5 1 1.5 2M k t B t d J 06 t J 07
15
Market Beta m easu red June06 to June07
Ex Ante Firm-level Risk (ES) and Performance During Crisis
UBHCBK
.5c0
8
Y = -0.29 – 6.73 * ES + error, adj R2=2.44% (-2.80) (-1.88)
PBCT
B ERBR K
CBSSW FC
ATAGE
AFLCB
S AF
TRV
AOC
B RKCBH
BLKCG
BOT
MA
0ul
y07
to D
ec
TR OW
BR K
SN VL UK
CINF
LTR MTB
W FCMMC
J PMHU M
UNP
BK
N TRSTRV
PNCTMK PGR
U SB
SEIC
UNMBBT
STT
SC HW
A LL
NYBAMTD
BLK
MET A ET
FIS
W LPA IZ
FNF
ICE
MA
W U NMX
UNH5ng c
risis
: Ju
CMA
EV
FITBRF
BEN
HBA NCNALN C
MIAXPBAC CIKEY
LM
STI
SEIC
MSCVHHNT
COF
HIG
ZIONGS
JNS
W LP
P FGPRU
CME
A IZ
AMP N YX
UNH-.5R
etur
n du
ri
SOVW B
FNM
ME R
NCCCFCSLM
AIGB SC
CMBI
FRE
HNT
AB K LEHW M
HIG
ETFCACAS CITGNW CBG
-1
.01 .02 .03 .04 .05 .06
16
ES measured June06 to June07
Ranking by systemic risk (MES) as of June 2007
MES (per share & $)
• Top 4 out of 10 firms ranked by MES pre-crisis have effectively failed (Bear, Lehman, ML, CIT), two received government support (GS, MS), others interesting (eTrade, CBRE, Charles Schwab, Ameritrade)• On a dollar basis for MES or insurance costs, Citigroup and JPMorgan also enter top 10
Ranking as of June 07 MES (%) MES ($)
E TRADE FINANCIAL CORP 1 37
BEAR STEARNS COMPANIES INC 2 20
C B RICHARD ELLIS GROUP INC 3 54
LEHMAN BROTHERS HOLDINGS INC 4 12LEHMAN BROTHERS HOLDINGS INC 4 12
MORGAN STANLEY DEAN WITTER & CO 5 4
GOLDMAN SACHS GROUP INC 6 5
MERRILL LYNCH & CO INC 7 6
SCHWAB CHARLES CORP NEW 8 16
C I T GROUP INC NEW 9 50
T D AMERITRADE HOLDING CORP 10 42
T ROWE PRICE GROUP INC 11 36
EDWARDS A G INC 12 68
FEDERAL NATIONAL MORTGAGE ASSN 13 8
JANUS CAP GROUP INC 14 76
FRANKLIN RESOURCES INC 15 13
LEGG MASON INC 16 44
AMERICAN CAPITAL STRATEGIES LTD 17 62
STATE STREET CORP 18 24
COUNTRYWIDE FINANCIAL CORP 19 27
17
COUNTRYWIDE FINANCIAL CORP 19 27
EATON VANCE CORP 20 75
II. Predictive power: Horse-race (Table 2)
Panel A
The dependent variable is Event return, the company stock returns during the crisis
Model 1 Model 2 Model 3 Model 4 Model 5 Model 6Intercept -0.29***
(-2.80) -0.23** (-2.45)
-0.20** (-2.24)
-0.22** (-2.43)
-0.23** (-2.22)
-0.22** (-2.05)
ES -6.73* (-1.88)
2.03 (0.37)
-0.29 (-0.05)
Beta -0.24*** (-2.77)
0.08 (0.32)
0.04 (0.13)
MES5 16 78*** 23 27*MES5 -16.78(-3.26)
-23.27(-1.73)
MES10 -20.42*** (-3.08)
-22.74 (-1.25)
Adj. R2 2.44% 6.19% 8.70% 7.75% 7.22% 5.88%
18
Predictive power: Lead (Table 2)
Panel B
The dependent variable is Event return, the company stock returns during the crisis. The explanatory variable is MES5 as computed over different pre-crisis periods.
June06 June06 May06 Apr06 Mar06 Jan06 June06-June07
June06-May07
May06-Apr07
Apr06-Mar07
Mar06-Feb07
Jan06-Dec06
Intercept -0.20**
(-2.24)-0.26*** (-3.17)
-0.32*** (-3.86)
-0.32*** (-3.86)
-0.39*** (-5.41)
-0.38*** (-5.48)
MES5 -16.78*** (-3.26)
-13.92*** (-2.92)
-8.62* (-1.97)
-8.62* (-1.97)
-5.43 (-1.27)
-6.55 (-1.46)( 3.26) ( 2.92) ( 1.97) ( 1.97) ( 1.27) ( 1.46)
Adj. R2 8.70% 6.92% 2.76% 2.76% 0.60% 1.12%
Panel C The dependent variable is Event return, the company stock returns during the crisis. The explanatory variable is MES10 as computed over different pre-crisis periods.
June06-J 07
June06-M 07
May06-A 07
Apr06-M 07
Mar06-F b07
Jan06-D 06June07 May07 Apr07 Mar07 Feb07 Dec06
Intercept -0.22**
(-2.43)-0.24*** (-2.91)
-0.30*** (-3.63)
-0.30*** (-3.56)
-0.35*** (-4.39)
-0.39*** (-5.72)
MES10 -20.42*** (-3.08)
-18.70*** (-2.97)
-13.47** (-2.26)
-13.30** (-2.30)
-10.24 (-1.65)
-7.67 (-1.45)
Adj R2 7 75% 7 18% 3 90% 4 07% 1 67% 1 08%
19
Adj. R 7.75% 7.18% 3.90% 4.07% 1.67% 1.08%
III. Cross-sectional variation across types of financial institutions
Based on two-digit SIC codes:
• 60 = Depository Institutions (JPMorgan, Citigroup, WAMU,…)
• 61 = Non-depository Institutions (Fannie Freddie AMEX Mastercard )61 Non depository Institutions (Fannie, Freddie, AMEX, Mastercard,…)
+ 62 other than 6211 (CBOT, CME, etc.)
+ 65 = Real estate
+ 67 = Holding and Other Investment Offices (Fifth Third, NYSE Euronext, Blackrock, …)
• 63 = Insurance Carriers (AIG, Berkshire Hathaway, Countrywide,…)
+ 64 = Insurance Agents, Brokers, Service (Metlife, Hartford Financial, …)
• 6211 = Security and Commodity Brokers (Goldman Sachs, Morgan Stanley,…)
Summary of results:
• Across categories, Security and Commodity Brokers have the highest systemic risk
• Within institution type some evidence that systemic risk is higher for (results not yet reported)• Within institution type, some evidence that systemic risk is higher for (results not yet reported)- Larger size
– Lower capital to assets ratio
– Higher debt to assets ratio
20
– Higher short-term debt to assets ratio (SIC code 6211 = Security brokers and dealers, 6221 = Commodity contractsbrokers and dealers)
Systemic risk of different types of institutions (Table 3)
Panel A
Descriptive statistics of the measures Event return, beta, ES, MES5 and MES10.
Depository Institutions, SIC code=60. EventReturn ES Beta MES5 MES10
Average -43.16% 2.23% 0.87 1.42% 1.12% Median -41.07% 2.11% 0.82 1.31% 1.09%
Std. Dev. 35.40% 0.48% 0.19 0.34% 0.25% Min -99 61% 1 27% 0 53 0 88% 0 66%Min 99.61% 1.27% 0.53 0.88% 0.66%Max 35.63% 3.58% 1.33 2.12% 1.71%
N 29 29 29 29 29
Panel B Descriptive statistics of the measures Event return, beta, ES, MES5 and MES10.
Other: Non-depository Institutions etc. SIC code=61, 62(except 6211), 65, 67. EventReturn ES Beta MES5 MES10
Average -52.29% 3.35% 1.22 1.92% 1.48% Median -57.90% 3.17% 1.18 1.83% 1.46%
Std. Dev. 32.26% 1.06% 0.35 0.63% 0.51%Min -98.78% 1.79% 0.67 0.92% 0.31% Max 10.12% 5.82% 2.10 3.36% 2.40%
N 27 27 27 27 27
21
Systemic risk of different types of institutions (Table 3)
Panel C
Descriptive statistics of the measures Event return, beta, ES, MES5 and MES10.
Insurance: SIC code=63 and 64. EventReturn ES Beta MES5 MES10
Average -43.78% 2.44% 0.78 1.28% 0.98% Median -43.84% 2.29% 0.76 1.38% 0.98%
Std. Dev. 32.26% 0.69% 0.23 0.39% 0.30% Min -98.47% 1.39% 0.34 0.39% 0.36% Max 13 56% 4 42% 1 51 2 09% 1 67%Max 13.56% 4.42% 1.51 2.09% 1.67%
N 36 36 36 36 36
Panel D Descriptive statistics of the measures Event return, beta, ES, MES5 and MES10.
Security and Commodity Brokers: SIC code=6211. EventReturn ES Beta MES5 MES10
Average -59.09% 3.61% 1.61 2.68% 2.04% Median -68.40% 3.46% 1.60 2.64% 2.10%
Std. Dev. 36.23% 0.68% 0.24 0.34% 0.35%Min -99.82% 2.88% 1.21 2.26% 1.38% Max -0.71% 5.24% 1.96 3.29% 2.51%
N 10 10 10 10 10
22
Effect of institution type on systemic risk (Table 4)
Fixed Effects Realized
return in the crisis
MES5 Beta
Depository Institutions
-0.43***(-6.93)
0.01***(17.07)
0.87***(18.12)
Other -0.52*** (-8.10)
0.02*** (22.17)
1.22*** (24.66)
Insurance -0.44*** (-7.83)
0.01*** (17.11)
0.78*** (18.31)
(
Security and Commodity
Brokers
-0.59*** (-5.57)
0.03*** (18.88)
1.61*** (19.76)
Adj. R2 66.42% 93.33% 94.21%
In all columns, all coefficients save for the first and third are statistically significantly different from each other.
Pre-crisis systemic risk contribution (MES5 * Size)
Expected Shortfall (in Billion Dollars) Data throughJune 07
0 1 2 3 4 50 1 2 3 4 5
CJPMBACMSGSGS
MERWFCFNMAIGWBAXPLEHBENMETWMSCHPRU
BKBKNYXFREBSCUSBUNHUNP
24
UNPSTT
NOTE: Systemic risk (MES) is NOT just Size!
ICE
.04
FNMAMTD
GS
JNSCIT
CBGAMP
ICE
NYX
TROW
MER
AGE
BSC
MSSCHW
LEH
ETFC
NMX
.03
SOV
PBCT SNV UB
CMA
RFMTB
WBWFCHBAN
JPMBK
MI
NCC
NTRS
BACPNCKEY
C
BBT
STT
WM
CBHHCBK
WU
LUK
CBSS
EV
FITB
BEN
UNP
FNM
AXP
LMSEIC
FRECOF
ACAS
BLK
JNS
FISCME
BOT
MA
BER
CINF
LTR
CNAHUM
LNCCBSAF
TRVAOCTMK
CIPGR
CFC
UNM
MBI
ABK HIG METAET
PFG
PRUCG
AIZ
GNW
TROWAGE
.02
ME
S5
PBCT SNV MI
USBSTI
NYBZION
CBSSATSLM
BRK
CINF
MMC AFL
AOCTMK
AIG
CVHHNT ALL
BRK
WLP
AIZFNF
UNH
.01
0
6 8 10 12 14Log Total Assets
Y = 0.02 - 0.0003 * log AT + error, adj R2 = -0.32% (5 08) ( 82)
25
(5.08) (-.82)
IV. Cyclical properties and stability over time
Time-series analysis:
Systemic risk measures rise as crises approach a crucial feature for our proposal to follow Systemic risk measures rise as crises approach, a crucial feature for our proposal to follow
Beta: Appears more variable over time; Picks up many recessions; Not always coincident with market stresses
MES: Appears more stable; Picks up recessions as well as market stresses (e.g., 1987 stock market crash)
Important observation: Systemic risk of financial sector and recessions are NOT always coincident
Systemic risk of Security Dealers and Brokers is robustly higher than that of others over 1963 to 2008
Likely a reflection of their (tail) risk-taking, higher leverage and financial fragility (short-term debt)
However, during stress periods, the gap in systemic risk between different groups appears to narrow
26
Systemic risk (MES5) over time for different groups
Annual Value-Weighted Marginal ES(5%) by groups.
0 11
0.09
0.11
0.05
0.07
0.01
0.03
-0.011963 1968 1973 1978 1983 1988 1993 1998 2003 2008
NBER Recession Depositories Others Insurance Sec. and Comm.
Systemic risk (MES5) over time for different groups
Annual Equally-Weighted Marginal ES(5%) by groups.
0 075
0.095
0.055
0.075
0.015
0.035
-0.0051963 1968 1973 1978 1983 1988 1993 1998 2003 2008
NBER Recession Depositories Others Insurance Sec. and Comm.
Systemic risk (beta) over time for different groups
Annual Value-Weighted Beta wrt. Market return by groups.
2 2
2.7
1.7
2.2
0.7
1.2
0.21963 1968 1973 1978 1983 1988 1993 1998 2003 2008
NBER Recession Depositories Others Insurance Sec and CommNBER Recession Depositories Others Insurance Sec. and Comm.
Systemic risk (beta) over time for different groups
Annual Equally-Weighted Beta wrt. Market return by groups.
3
2.5
3
1.5
2
0.5
1
01963 1968 1973 1978 1983 1988 1993 1998 2003 2008
NBER Recession Depositories Others Insurance Sec and CommNBER Recession Depositories Others Insurance Sec. and Comm.
Firm-level risk (ES) over time for different groups
Annual Value-Weighted ES(5%) by groups.
0.11
0.13
0.15
0.07
0.09
0.03
0.05
-0.01
0.01
1963 1968 1973 1978 1983 1988 1993 1998 2003 2008
NBER Recession Depositories Others Insurance Sec. and Comm.
Firm-level risk (ES) over time for different groups
Annual Equally-Weighted ES(5%) by groups.
0.11
0.13
0.15
0.07
0.09
0.03
0.05
-0.01
0.01
1963 1968 1973 1978 1983 1988 1993 1998 2003 2008
NBER Recession Depositories Others Insurance Sec. and Comm.
V. Predictive power of systemic risk measures for crises (Table 5)
MES f th fi i l t di t th li d h tf ll f th k t ith l dMES of the financial sector predicts the realized shortfall of the market with a one-year lead:
Dependent Variable is expected shortfall of the market at 5%
Independent variable Equally weighted market beta Value weighted market betap q y g gIntercept 0.017***
(5.39) 0.0162***
(4.81) 0.014***
(4.16) 0.001 (0.13)
-0.001 (-0.10)
-0.010 (-1.13)
Lag 1 0.005 (1.15)
0.0004 (0.06)
-0.002 (-0.29)
0.02141*** (2.99)
0.017* (1.90)
0.022** (2.33)
Lag 2 0.006 (0.91)
-0.002 (-0.33)
0.006 (0.63)
-0.008 (-0.71)
Lag 3 0.013** (2.08)
0.019** (2.04)
Adj. R2 0.40% 0.28% 4.39% 8.95% 7.75% 11.75%
Independent variable Equally weighted MES5 Value weighted MES5 Intercept 0.014***
(6.61) 0.013***
(5.86) 0.014***
(6.61) 0.013***
(5.86) 0.014***
(6.61) 0.013***
(5.86) Lag 1 0.357***
(3.03) 0.327***
(2.69) 0.357***
(3.03) 0.327***
(2.69) 0.357***
(3.03) 0.327***
(2.69) Lag 2 0 097 0 031 0 097 0 031 0 097 0 031Lag 2 0.097
(0.82) 0.031(0.24)
0.097(0.82)
0.031 (0.24)
0.097(0.82)
0.031(0.24)
Lag 3 0.151 (1.24)
0.151 (1.24)
0.151 (1.24)
Adj. R2 16.48% 17.05% 16.48% 17.05% 16.48% 17.05%
33
VI. Evidence for other crises (Details available upon request)
Early 80’s- Pre-event = 81-82; Event = 83-84
- MES works better than Beta
Savings and Loans- Pre-event = 87; Event = 88-89
N ith MES B t d d j b- Neither MES nor Beta do a good job
Late 90’s- Pre-event = Oct 96 to Sep 97; Event = Oct 97 till end of 98
- Beta does better than MES
In ALL cases, ES does the worst compared to systemic or systematic risk measures
34
Advantages of MES and caveats
Does not rely on a particular “percentile” (it looks at the whole tail)
• This is especially helpful when one needs to condition on a tail event
• Calculation straightforwardCalculation straightforward
Flexible technology
• Can be done for profits, credit losses, etc.
• Break down by divisions, desks, assets, geographical regions
• Consistent with M&As, changes in size, positions, etc.
Caveats on statistical methods
• Cyclical behavior: Can capture systemic risk, but may be just systematic risk (beta)
• Past data vs. future crisis
Complement with scenario analysis
Qualitative inputs: Inter-connectedness, complexity, concentration (e.g., JPMorgan Chase)
35
Q p , p y, ( g , g )
VII. Our proposal for regulating systemic risk
1. Systemic Capital Requirement (Basel III)
• Capital requirement proportional to estimated systemic risk
2. Systemic Fees (FDIC-style)
• Fees proportional to estimated systemic risk
• Create systemic fund.
3. Our preferred approach: Systemic Insurance provided by the private/publicp pp y p y p p
• Compulsory insurance of each bank’s own losses during general crisis
• Payment goes to systemic fund, not the bank itself
M k i f i b f h i b h f h• Market price of insurance, but most of the insurance bought from the government
– Not enough capital for provision of ALL systemic insurance (Problems with Monolines, A.I.G.,…)
– Analogy to terrorism reinsurance by the government (TRIA, 2002)
Advantages of our proposal
• Incentives to “organically” limit systemic risk (e.g., to lower short-term debt, correlation with market)
• Estimates of systemic risk (by regulator and by the insurance market)
36
• Reduce risk and cost of bailout (systemic fund)
Specific design questions
What should be the maturity at the time of each purchase of insurance?
• 5-year (should roughly cover expected time between cycles)
How frequently should the insurance be acquired?
• Monthly to reflect changing risk in new insurance purchase, so rolling 1/60th of insurance purchased each month
Should the price charged by the regulator be the same as that charged by the private sector?
• Perhaps the same, but there could be some discounting as funds when deployed for resolution also earn a return
How should the private insurers’ counterparty risk be mitigated?
• Well-capitalized; Perhaps even hold 100% capital against the largest insurance they have provided; And perhapseven trade on centralized clearinghouse/exchange…
• Buyers of insurance will have to pay for capitalization of these insurance providers
What proportion of insurance provision should be from the private sector?
• Need well-capitalized but also competitive insurance provision
When do institutions “fail”? How should the regulator deploy the insurance premia and payments collected?
• Pre-announced early intervention points (as in FDIC’s Prompt Corrective Action) for not requiring any furtherinsurance purchase, suspension of dividend payments, receivership, …
37
p , p p y , p,
• Augment the systemic risk fund; Use reserves for resolution of systemic failures (guarantees, bailouts, etc.)
Regulating systemic risk: Pros and cons of various approaches
C it l T P i t I P bli /P i tCapital requirements
Taxes Private Insurance Public/Private Insurance
Advantages Consistent with existing regulations
Transparent and easy to implement
Easy to adjustCreate a systemic
fund
No need for extra capital on BS
Extract market prices
Market pricePublic power
Disadvantages Cost of keeping large capital on balance sheet
Hard to figure out the price
Market not large enough for real systemic risk
LOLR still there
GovernanceCoordination
Find correct public price
38
Illustration of how systemic risk insurance might be priced
G l Goals
• Illustrate that our proposal is not just an abstract notion but can be readily implemented
• To provide a rough estimate of the price of systemic risk insurance and study its determinants
Assumptions: Multi-variate normality, representative agent, constant relative risk-aversion (CRRA)
• Based on Stapleton and Subrahmanyam (1984)
Additional calculations (To be completed)
• Assess the impact of buying systemic risk insurance on earnings and P/E ratios of banks
• Construct and understand time-series of insurance costs for systemically important institutions
• Understand the relative costs across institutions
NOTE: Under market pricing of insurance
• Actual prices may not be based on multi normal distributions• Actual prices may not be based on multi-normal distributions
• Market price will factor in non-normal distributions, conditionality of distributions in market tail events,time-varying risk premium, etc.– Key benefit of private part of public-private scheme
39
– Analogy: Black-Scholes model versus Market prices of options
Outline
Th ti l l l ti Theoretical calculation
• Specification of the systemic risk insurance
• Valuation
Illustrative calculations
• Base case parameters
• 3D graphs of the effect of volatility and correlation parameters
Calculations of financial firm insurance charges
• Description of the assumptions and model runs
• Figures presenting most systemic firms, 2004-2007
40
Specification of the systemic risk insurance
Assume that the financial institution is required to take out insurance on systemic losses tied to the marketvalue of equity of the firm and the overall sector.
A systemic risk insurance is defined by:
• the market value of the equity of the aggregate financial sector falling below a threshold, and
• the required payment at maturity of the claim, which is the difference between some pre-specified marketvalue of the equity of the financial institution and its actual market value.
The payoff at maturity T can be represented mathematically as The payoff at maturity T can be represented mathematically as
max( ,0) max( ,0)S MTM
iS MTM
K SS iTK S K S
Question: What is its price assuming joint log-normality of the two equity returns?
Illustrative calculation:
• Market volatility of 20%, bank volatility of 50% and correlation of 50%
• The price of ensuring bank value beyond 50% loss when market has made 50% loss is of the order of$5.5bln on a $1trn notional [Black-Scholes on bank = $66bln, Black-Scholes on market = $1.5bln]
• Price can however rise dramatically when there is a “perfect storm”
41
Price can however rise dramatically when there is a perfect storm
Solution to the valuation problem
)0max()0,max(1 SK dSdSSSSKV MTMS
)(
, )0,max(
1
00KK
iTMTiTMTiTSSKrt
dSdSSSSK
dSdSSSSKV
iSMS
iMTMStT
,,)(00
iTMTiTMTiTSrdSdSSSSK
itT
Where,
SrtTSSrtTS
SStTiTMT
SitTitiT
MStTMtMT
TMi
iTMTMiiSMSeSS
2
2)(2
2)(
)21(21
2lnln)(ln
2lnln)(ln
)1()(21,
tT
SrtTS
tT
SrtTS
Mi
tT
StS
tTT
SitTitiT
MStTMtMT
iS
itiT
MS
MtMT
2
2)(2
2)(
22
lnln)(lnlnln)(ln
)(
2
tTtTMiiSMS
Base Case Parameters
Initial bank asset value: 10itS Initial market asset value:
Bank crisis boundary:
it10MtS
5iSK
Market crisis boundary:
Time-to-maturity:5
MSK
4 tT Risk-free rate:
Bank volatility: %50iS
%4r
Market volatility:
Correlation:
%20MS
5.0Mi
Joint effects of correlation and bank volatility
Joint effects of correlation and market volatility
Joint effects of bank and market volatility
Description of Financial Firm Insurance Charges
A ti Assumptions
• Multivariate normality– Each year, 2004-2007, take the prior year’s volatility and correlation of the equity of the financial firms and the
market.
• Payoff based on (i) an event trigger of a 40% drop in the stock market, and (ii) the difference between astrike value of equity such that (market value of equity/total assets)=10% and the current equity value ofthe firm. (We also tried 25% drop in the market, and strikes based on 5% and 2.5%.)
• 4-year maturity and current 1-year treasury rate.
Caveats
• Strong evidence that multivariate normality does not hold, in particular, equity returns have fat tails andare more correlated during crises. This would greatly increase the insurance cost though not necessarilychange the ranking across firms.
• We treat the liabilities of insurance companies, investment banks and commercial banks on the samelevel in terms of measuring the insurance payoff. It is desirable to perhaps treat different types offinancial institutions differently.
Illustrative calculations (40% drop, 10% equity/assets)us a ve ca cu a o s ( 0% d op, 0% equ y/asse s)
• Time-series of insurance charges of top 10 systemic firms (based on 2007) over the 2004-2007 period.
• Tables of systemic firm ranking based on insurance charges over the period 2004-2007 as a function of $charges and of $ charges as a % of equity value.
47
Illustrative insurance charges of the top 10 systemic firms in 2007 by $mm (2004-2007)
Note that the insurance charges are highest in 2004, reflecting the higher volatility during this
600
Note that the insurance charges are highest in 2004, reflecting the higher volatility during this period. In this sample, volatility was highest at the beginning of the boom and lowest going into the bust, producing countercyclical insurance charges.
500MORGAN STANLEY DEAN WITTER & COCITIGROUPINC
300
400CITIGROUP INCMERRILL LYNCH & CO INCJPMORGAN CHASE & COGOLDMAN SACHS GROUP INCFEDERALHOMELOANMORTGAGECORP
100
200
FEDERAL HOME LOAN MORTGAGE CORPFEDERAL NATIONAL MORTGAGE ASSNLEHMAN BROTHERS HOLDINGS INCBEAR STEARNS COMPANIES INCMETLIFE INC
0
100
2004 2005 2006 2007
BANK OF AMERICA CORP
2004 2005 2006 2007
Rank 2004 2005 2006 2007
1FEDERAL NATIONAL MORTGAGE ASSN
FEDERAL NATIONAL MORTGAGE ASSN
MORGAN STANLEY DEAN WITTER & CO
MORGAN STANLEY DEAN WITTER & CO
2MORGAN STANLEY DEAN WITTER & CO
MORGAN STANLEY DEAN WITTER & CO
FEDERAL NATIONAL MORTGAGE ASSN CITIGROUP INC2WITTER & CO & CO ASSN CITIGROUP INC
3 JPMORGAN CHASE & COFEDERAL HOME LOAN MORTGAGE CORP GOLDMAN SACHS GROUP INC MERRILL LYNCH & CO INC
4MERRILL LYNCH & CO INC JPMORGAN CHASE & CO MERRILL LYNCH & CO INC JPMORGAN CHASE & CO
5GOLDMAN SACHS GROUP INC MERRILL LYNCH & CO INC JPMORGAN CHASE & CO GOLDMAN SACHS GROUP INCLEHMAN BROTHERS HOLDINGS LEHMAN BROTHERS HOLDINGS FEDERAL HOME LOAN
6LEHMAN BROTHERS HOLDINGS INC GOLDMAN SACHS GROUP INC
LEHMAN BROTHERS HOLDINGS INC
FEDERAL HOME LOAN MORTGAGE CORP
7PRUDENTIAL FINANCIAL INCLEHMAN BROTHERS HOLDINGS INC METLIFE INC
FEDERAL NATIONAL MORTGAGE ASSN
8CITIGROUP INC PRUDENTIAL FINANCIAL INC BEAR STEARNS COMPANIES INCLEHMAN BROTHERS HOLDINGS INC
BEAR STEARNS COMPANIES9BEAR STEARNS COMPANIES INC METLIFE INC PRUDENTIAL FINANCIAL INC BEAR STEARNS COMPANIES INC
10METLIFE INC CITIGROUP INCHARTFORD FINANCIAL SVCS GROUP I METLIFE INC
11HARTFORD FINANCIAL SVCS GROUP I BEAR STEARNS COMPANIES INC CITIGROUP INC BANK OF AMERICA CORP
12BANK OF AMERICA CORP BANK OF AMERICA CORP BANK OF AMERICA CORP PRUDENTIAL FINANCIAL INC
13WACHOVIA CORP 2ND NEWAMERICAN INTERNATIONAL GROUP IN WASHINGTON MUTUAL INC
HARTFORD FINANCIAL SVCS GROUP I
14WASHINGTON MUTUAL INCHARTFORD FINANCIAL SVCS GROUP I COUNTRYWIDE FINANCIAL CORP COUNTRYWIDE FINANCIAL CORP
15LINCOLN NATIONAL CORP IN WACHOVIA CORP 2ND NEW WACHOVIA CORP 2ND NEW WACHOVIA CORP 2ND NEW15LINCOLN NATIONAL CORP IN WACHOVIA CORP 2ND NEW WACHOVIA CORP 2ND NEW WACHOVIA CORP 2ND NEW
RANKINGS of MOST SYSTEMIC FINANCIAL INSTITUTIONS BY HYPOTHETICAL $ INSURANCEC A G S f 2004 200CHARGES from 2004-2007
Rank 2004 2005 2006 20071BEAR STEARNS COMPANIES INC BEAR STEARNS COMPANIES INC BEAR STEARNS COMPANIES INC BEAR STEARNS COMPANIES INC
2GENWORTH FINANCIAL INCFEDERAL HOME LOAN MORTGAGE CORP
FEDERAL NATIONAL MORTGAGE ASSN
FEDERAL HOME LOAN MORTGAGE CORP
LEHMAN BROTHERS HOLDINGS FEDERAL NATIONAL MORTGAGE MORGAN STANLEY DEAN WITTER LEHMAN BROTHERS HOLDINGS3LEHMAN BROTHERS HOLDINGS INC
FEDERAL NATIONAL MORTGAGE ASSN
MORGAN STANLEY DEAN WITTER & CO
LEHMAN BROTHERS HOLDINGS INC
4PRUDENTIAL FINANCIAL INCMORGAN STANLEY DEAN WITTER & CO
LEHMAN BROTHERS HOLDINGS INC MERRILL LYNCH & CO INC
5MORGAN STANLEY DEAN WITTER & CO LINCOLN NATIONAL CORP IN GOLDMAN SACHS GROUP INC
MORGAN STANLEY DEAN WITTER & CO
6LINCOLN NATIONAL CORP INLEHMAN BROTHERS HOLDINGS INC MERRILL LYNCH & CO INC
FEDERAL NATIONAL MORTGAGE ASSN
7FEDERAL NATIONAL MORTGAGE ASSN GOLDMAN SACHS GROUP INC METLIFE INC GOLDMAN SACHS GROUP INC
8HARTFORD FINANCIAL SVCS GROUP I MERRILL LYNCH & CO INC
HARTFORD FINANCIAL SVCS GROUP I COUNTRYWIDE FINANCIAL CORP8GROUP I MERRILL LYNCH & CO INC GROUP I COUNTRYWIDE FINANCIAL CORP
9METLIFE INCHARTFORD FINANCIAL SVCS GROUP I PRUDENTIAL FINANCIAL INC METLIFE INC
10MERRILL LYNCH & CO INC PRUDENTIAL FINANCIAL INC LINCOLN NATIONAL CORP INHARTFORD FINANCIAL SVCS GROUP I
11GOLDMAN SACHS GROUP INC GENWORTH FINANCIAL INC AMERIPRISE FINANCIAL INCPRINCIPAL FINANCIAL GROUP INC11GOLDMAN SACHS GROUP INC GENWORTH FINANCIAL INC AMERIPRISE FINANCIAL INC INC
12 JPMORGAN CHASE & CO METLIFE INC COUNTRYWIDE FINANCIAL CORP LINCOLN NATIONAL CORP IN
13PRINCIPAL FINANCIAL GROUP INC PRINCIPAL FINANCIAL GROUP INC JPMORGAN CHASE & CO PRUDENTIAL FINANCIAL INC
14E TRADE FINANCIAL CORP JPMORGAN CHASE & CO UNUM GROUP JPMORGAN CHASE & CO
15UNUM GROUP E TRADE FINANCIAL CORP SOVEREIGN BANCORP INC CITIGROUP INC15UNUM GROUP E TRADE FINANCIAL CORP SOVEREIGN BANCORP INC CITIGROUP INC
16TRAVELERS COMPANIES INC UNUM GROUP PRINCIPAL FINANCIAL GROUP INC AMERIPRISE FINANCIAL INC
17C I G N A CORP WASHINGTON MUTUAL INC E TRADE FINANCIAL CORP E TRADE FINANCIAL CORP
18SOVEREIGN BANCORP INC C N A FINANCIAL CORP WASHINGTON MUTUAL INC C I T GROUP INC NEW
19WASHINGTON MUTUAL INC COUNTRYWIDE FINANCIAL CORP COMMERCE BANCORP INC NJ WASHINGTON MUTUAL INC
20COMMERCE BANCORP INC NJ COMMERCE BANCORP INC NJ HUNTINGTON BANCSHARES INC COMMERCE BANCORP INC NJ
RANKINGS of MOST SYSTEMIC FINANCIAL INSTITUTIONS BY HYPOTHETICAL $ INSURANCE CHARGES from 2004-2007 as a % of EQUITY
Conclusion
Departure needed from regulation that focuses solely on institution-level risk
• Not only does it not serve its purpose, it is in fact likely to do harm!
It’s time to quantify systemic risk of financial institutions
• Marginal Expected Shortfall offers one measure
• Robustness of different systemic risk measures should be evaluated carefullyy y
• Confidence in measures would lead to eventual pricing/taxing of systemic risk contributions
S i i k CAN b d d i i k i CAN b i d d bl i Systemic risk CAN be measured and systemic risk insurance CAN be priced under reasonable assumptions
Private participation in measurement and pricing of systemic risk likely to be beneficial to regulators
• Market scrutiny of systemic risk
• Price discovery of systemic risk insurance given likely non-linearities
• Regulators can build in an automatic reinsurance through our public-cum-private scheme
51