with Yuliy Sannikov M arkus K Brunnermeier TexPoint fonts used in EMF Read the TexPoint manual before you delete this box A A A A A A 1 CoVaR with Tobias Adrian ID: 911043
Download Presentation The PPT/PDF document "M acroModel with Financial Sector" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
Slide1
MacroModel with Financial Sector with Yuliy Sannikov Markus K. Brunnermeier
TexPoint fonts used in EMF. Read the TexPoint manual before you delete this box.: AAAAAA
1
CoVaR
with Tobias Adrian
Slide2Traditional BankingRole of banks2
Originate & distributeSecuritization
Pooling
Tranching
Insuring
(CDS)
Dual purpose
Tradable assetCollateral feeds repo market for levering
Channel fundsLong-run repaymentProspect of selling offMaturity transformationRetail fundingWholesale funding (money market funds, repo partners, conduits, SIVs, …)Info-insensitive securitiesDemand depositsABCP, MTN, overnight repos, securities lending
Demand
deposits
A
L
Loans
(long-term)
Equity
ABCP/MTN
AAA
Loans
(long-term)
Equity
BBB
…
SIV/Conduit
s
Shadow banking system
Slide3Changing banking landscapeTraditional BankingRole of banks3
Originate & distributeSecuritization
Pooling
Tranching
Insuring
(CDS)
Dual purpose
Tradable assetCollateral feeds repo market for levering
Channel fundsLong-run repaymentProspect of selling offMaturity transformationRetail fundingWholesale funding (money market funds, repo partners, conduits, SIVs, …)Info-insensitive securitiesDemand depositsABCP, MTN, overnight repos, securities lending
Demand
deposits
A
L
Loans
(long-term)
Equity
ABCP/MTN
AAA
Loans
(long-term)
Equity
BBB
…
SIV/Conduit
Slide4Two questionsWhy is reaction so sharp?Liquidity spirals(non-linear dynamics due to adverse feedback loop)Real economic effectsmacro model with financial sector (w/ Sannikov)Too much leverage and maturity mismatch?Identify and measure externalities(risk spillovers rather risk of a bank in isolation)CoVaR (w/ Adrian)
4ABS issuance
Source: JPMorgan
Slide5OverviewTheory (with Sannikov)Spirals: Non-linear adverse feedback loops + volatility effectsExternalitiesImplementation (with Adrian)CoVaR: measuring systemic risk contribution/externalitiesOne method: Quantile regressionsAddressing procyclicalites due to spirals5
Slide6Brunnermeier-Sannikov (new)
Entrepreneurs
Needs financing
Start projects (trees with
payoff
a
t
K
t)dat/at=gdt +σdZt capital depreciatesinvestment6 Financial Experts
Monitoring (growth of at)Securitizes “trees”
to expand investment
Households
Provide financing
D
E
B
T
EQU
ITY
A
L
outside
inside
Optimal
dynamic
contract
direct financing
a
t
grows slower
Slide7Role of financial experts in the economy: Entrepreneurs
7
Financial Experts
Households
Securitize investments
and sell products to
households
Direct financing
Provide funding moreefficiently than direct lending Both entrepreneurs and households benefit from the financial sector
Slide8Model setup - overview
Entrepreneurs
Needs financing
Start projects (trees with
payoff
a
t
K
t)dat/at=gdt +σdZt capital depreciatesinvestment8 Financial Experts
Monitoring (growth of at)Securitizes “trees”
to expand investment
Households
Provide financing
D
E
B
T
EQU
ITY
A
L
outside
inside
Optimal
dynamic
contract
p
σ
p
Model setup - overview
Entrepreneurs
Needs financing
Start projects (trees with
payoff
a
t
K
t)dat/at=gdt +σdZt capital depreciatesinvestment9 Financial Experts
Monitoring (growth of at)Securitizes “trees”
to expand investment
Households
Provide financing
D
E
B
T
EQU
ITY
A
L
outside
inside
Optimal
dynamic
contract
direct financing
a
t
grows slower
Slide10Model setup: production of output and capitalProduction of output (numeraire) (apples): Yt = at Kt, where dat = a
tdZt Note everything will be scale invariant w.r.t. YtProduction of capital Kt
(trees):
dKt = (Φ
(I
t
/
Y
t
) - ) Kt dt,Investment-ratio It/Yt depends on ytνt price of capital (value of asset/tree)i.e. νt = price-earnings ratio dνt= tνdt+tνdZt(price of a tree in terms of apples divided by yt) max i {νtatKt
Φ(it/yt) – it} supply of capital: κ(νt)Kt 10
Slide11Financing of capitalAll agentsrisk neutral (for now) common discount rate of Direct financing through households (fraction 1-ψt)
Growth of a is zeroνt ≥ 1/(+)Break even for HH
1-(+
)νt +
t
ν
+
tν = 0 (HH)Indirect financing through financial experts (fraction ψ)Growth of a is g at cost b per y dat = gatdt+atdZt Reduced form for better resource allocation(monitor, service mortgage, channel continuation funding)νt ≤ (1-b)/(+-g)
11Capital appreciationearningsfinancing cost
Slide12Financing friction – optimal contractingExpert’s incentive problemeffort choice, yt are non-contractable, but asset value νtyt
is Incentive (“skin in the game”) constraint: tgνt - b 0
t = b/(gνt
)
Evolution of experts balance sheet
Asset side –
long maturity
d(
νtyt) = yt (tν+(g - )νt+ tν)dt + yt (νt + tν)dZtLiability sideDebt – overnight maturity
Outside equity: 1-tInside equitydnt = nt + yt [(1-b-(+-g)νt +
tν+ tν
)dt +
t (νt
+tν) dZ
t]
12
Slide13Competitive equilibriumState variable: t = Nt/Yt dt = …. (from Ito)Nt is aggregate wealth (net-worth) of financial expertsYt is aggregate output (scale-invariant)
Solve in terms ofP/E ratio: νt = ν(t)Expert i’s
value function: ftn
t = f(
t
)
n
t
Fraction of indirect investment:
ψ(t) ≤ 1Expert i’s Bellman equationntft dt = maxy E[d(ntft)] = maxy {(nt+y[1-b-(+-g)νt+
tν +tν]) ft+nt μtf+tfy
t(ν
t+tν
)} dt FOC: [1-b-(
+-g)ν
t+ tν +
tν] + t
f
t (
νt + t
ν) = 0 (FS)Household
FOC: 1-(+)ν
t + tν+
tν =(<) 0 (HH)
13
precautionary motive
=
μ
t
n
=
σ
t
n
Slide14Competitive equilibriumState variable: t = Nt/Yt dt = …. (from Ito)Nt is aggregate wealth (net-worth) of financial expertsYt is aggregate output (scale-invariant)
Solve in terms ofP/E ratio: νt = ν(t)Expert i’s
value function: ftn
t= f(t
)
n
t
Fraction of indirect investment:
ψ
≤ 1Expert i’s Bellman equationntft dt = maxy E[d(ntft)] = maxy {(nt+y[1-b-(+-g)νt+tν +
tν]) ft+nt μtf+tfyt(ν
t+t
ν)} dt
FOC: [1-b-(+-g)νt
+ tν +
tν] +
tf
t (
νt +
tν) = 0 (FS
)Household FOC: 1-(+
)νt + t
ν+ tν = 0 (HH
)
14
precautionary motive
=
μ
t
n
=
σ
t
n
1
2
3
Slide15Competitive equilibriumState variable: t = Nt/Yt dt = …. (from Ito)Nt is aggregate wealth (net-worth) of financial expertsYt is aggregate output (scale-invariant)
Solve in terms ofP/E ratio: νt = ν(t)Expert i’s
value function: ftn
t= f(t
)
n
t
Fraction of indirect investment:
ψ
≤ 1Expert i’s Bellman equationntft dt = maxy E[d(ntft)] = maxy {(nt+y[1-b-(+-g)νt+tν +
tν]) ft+nt μtf+tfyt(ν
t+t
ν)} dt
FOC: [1-b-(+-g)νt
+ tν +
tν] +
tf
t (
νt +
tν) = 0 (FS
)Household FOC: 1-(+
)νt + t
ν+ tν = 0 (HH
)
15
precautionary motive
=
μ
t
n
=
σ
t
n
1
2
3
1
2
3
Slide16Function of experts’ networth/GDP16
Marginal value of a $
Leverage
P/E ratio
GDP growth
Slide17Results 1: Non-linear dynamics - spiralsFact 1: financial sectorincreases growth but may also increase volatilityFact 2: Loss spiralprice is more volatile, as experts approach regime when they fire-sale assetsFact 3: “Outside-equity spiral” t = b/(gνt)for low νt , fraction of outside equity shrinks (difficult to raise because agency problem gets worse)Fact 4: Leverage spiralLeverage: [v(
t)ψt – t]/[v(t)ψt
]Internal: Own risk management for ψt
<1External: Margin/haircuts
spiral (see Brunnermeier-Pedersen, 09)
17
Slide18Introducing margin/haircut constraintLevel of debt is limited byIncentive constraint (in aggregate)+ haircut/margin constraint Spirit: asset can only be sold with a delayHaircut is multiple of price volatility h( νt + tν)/ ν
t Main changesprice-earning ratios go down price volatility goes up as long as haircuts don't binding (especially near the point where haircuts start binding)goes down when haircuts bindExperts value function rises (externalities – later)Internal risk management is enforced
Fear of haircut constraint becomes binding
18
Slide19Vol. and leverage with haircut constraints19
Slide20Graphs with haircut constraints (red)20
Slide21Graphs with haircut constraints (red)21
Slide22Graphs with haircut constraints (red)22
Slide23Results 2: Externalities – welfare“Too much” leverage/maturity-mismatch due to externalities?Financial regulation should focus on externalitiesWithin the financial sectorBetween financial sector and real economy (entrepreneurs)Two forms of inefficiencies:Inefficient (pecuniary) externalities– regulatory correction for an instantDynamic externality - commitment problem within an institution23
Slide24Result 2: Externalities – welfareFocus within financial sectorEffect of one expert’s choice of y on value function of everybody: f() (1-b+( + -g)
νt+tν+
tν) + f’(
)
σ
t
η
t(pt+tν)- f’()2 g + (f’() + f()) ψt (dtν/dψt
+ dtν/dψt) + f’() [1-b+(+-g)
νt+
tν+t
ν] +
[f’’()
+ f’()]
t
t(ν
t +
tp) + [f’’()
+ 2f’(
)]t
ψtt
dtν
/dψt
.
24
Slide25Results 2: Externalities – welfareFocus within financial sectorEffect of one expert’s choice of y on value function of everybody: f() (1-b+( + -g
)νt+tν+
tν) + f’(
)
σ
t
η
t(pt+tν)- f’()2 g + (f’() + f()) ψt (dtν/dψt
+ dtν/dψt) + f’() [1-b+(+-g
)νt+
tν+
tν] +
[f’’()
+ f’()]
t
t(
νt +
tp) + [f’’(
) + 2f’(
)]t
ψt
t dtν
/dψt
.
25
Affects the drift of
η
, and impacts other experts
Affects the volatility of
η
, and impacts other experts
Affects economic growth, and impacts other experts
Zero in individual expert’s FOC
Slide26On externalities – welfare analysisFocus within financial sectorEffect of one expert’s choice of y on value function of everybody: f() (1-b+( + -g)
νt+tν+
tν) + f’(
)
σ
t
η
t(pt+tν)- f’()2 g + (f’() + f()) ψt (dtν/dψt
+ dtν/dψt) + f’() [1-b+(+-g
)νt+
tν+
tν] +
[f’’()
+ f’(
)]t
t(
νt +
tp) + [f’’(
) + 2f’(
)]t
ψt
t dt
ν /dψt
.
26
(+) economic growth good for experts
Zero in individual expert’s FOC
Effect on value of other expert’s assets, through prices
(-) profit causes
η
to grow, which hurts other experts
(+) effect on expected value of cash
(-) effect on expected value of assets
Fire-sale externality
Slide27Externalities27
Slide28Related LiteratureEnd borrowers’ financing frictionsBernanke-Gertler-(Gilchrist), …, MishkinKiyotaki-Moore Financial sector’s frictions – liquidity spiralsBrunnermeier-PedersenDiamond-Dybvig, Allen-Gale, …He-KrishnamurthyDynamic contractingDeMarzo-Fishman-Sannikov, …
28
Slide29Differences to Bernanke-Gertler-GilchristBGG“small” aggregate shocks around steady stateidiosyncratic shocks are essentialDefault and associated costly state verification is more likelyAsset prices are driven by default (verification cost) due to idiosyncratic riskExpert’s rent is always zero (?)No incentive to keep “dry powder” (liquidity) … (No Bellman equ.)
Countercyclical leverageEntrepreneur take on same position after drop in networthLeverage increases after drop in net-worthDebt vs. Equity No fire-sale externality
29
BruSan
Focus on (large) aggregate shocks
(idiosyncratic shocks not essential)
(no restriction to steady state)
Asset
price drops
due to fire salesExpert’s rent depends on state
tIncentive to keep “dry powder” (liquidity) …Procyclical leverage
Experts reduce position after drop in networth
Liquidity spirals
Securitization
(
debt, inside +
outside equity)
Fire-sale externality
(rationale for regulation)
Slide30Differences to Kiyotaki-MooreKM – (Kiyotaki version)Zero-prob. temporary shockPersistent (dynamic loss spiral)Amplified through collateral valueNon- vs. productive (leveraged) sectorDual role of durable assetProductionCollateral
Exogenous contractOne period contractDebt is limited by collateral valueDurable asset doesn’t depreciates (capital, fully)30
BruSan
Permanent
TFP shocks
Margin/haircut spiral (leverage)
Loss
spiral
Investment
through leveraged financial sector
Dual role of durable assetProductionSecuritizationOptimal contract
Dynamic contract
Debt is limited
due idiosyncratic risk and costly state verification
δ
-depreciation rate
Slide31Differences to He-KrishnamurthyHe-Krishnamurthy Endowment economyGDP growth is exogenously fixedNo physical investmentNo direct investment in risky asset by householdsLimited participation modelContractingOnly short-run relationship (t to t+dt
) Fraction of return, fee Asset composition (risky vs. risk-free) is not contractableNon-effort lowers return by xdtx is
exogenous,not linked to fundamental Private benefit from shirking
No benchmarkingPricing Implications
When experts wealth declines, their market power increases, and so does their fee
Price impact depends on assumption that household have larger discount rate than experts
Procyclical
Leverage
In H-K calibration paper
No fee, households are rationed in their investmentAs expert wealth approaches 0, interest rate can go to –∞ Heterogeneous labor income for newborns of lDtNon-log utility function31BruSanProduction economy
GDP growth depends on net-wealthPhysical investmentDirect investments by all householdsContracting
(Potential) long-run relationship
Fraction of return, fee, size of asset pool
Effort increases
fundamental
growth to
gdt
Monetary
benefit from shirking
No benchmarking
Pricing
Implication
Price drop with state variable
Countercyclcial
Leverage
Entrepreneur take on same position after drop in
networth
Leverage increases after drop in net-worth
Slide32ConclusionIncorporate financial sector in macromodelHigher growthHigher volatilityMain insights:Adverse feedback loopsExternalities (rationale for financial regulation)Within financial sectorToward the real economy32
Slide33Macro-prudential regulationExternalities – “stability is a public good”Fire-sale externality Fire-sales depress prices also for othersVolatility: Precautionary hoarding uncertainty about future funding…Network externality: Hiding owns’ commitmentUncertainty for counterparties (of counterparties …)Countercyclical regulation – counteract spiralsRegulation strict during boomsLean against credit bubblesIncorporate funding structure
33
Slide34OverviewTheory (with Sannikov)Spirals: Non-linear adverse feedback loopsExternalitiesImplementation (with Adrian)CoVaR: measuring systemic risk contribution/externalitiesOne method: Quantile regressionsAddressing procyclicalites due to spirals34
Slide35“CoVaR” with Tobias AdrianSystemic risk measureCapture externalities and contribution to systemic risk“Clone property”Splitting one institution to 10 identical clones (which perfectly comove with each other) does not reduce systemic riskContrast to current regulation focus on risk in isolation, VaRincentive to hang on to others, become big, interconnectedprocyclicalAmplify non-linearities even further
35
VaR
1%
Slide36Who should be regulated?“Clone Property”Split individually systemic institution i in 10 identical clones c: CoVaRi= 10 CoVaRc
36groupexamples
macro-prudentialmicro-prudential
“individually systemic”
International banks
(national champions)
Yes
Yes
“systemic as part
of a herd”Leveraged hedge fundsYesNonon-systemic largePension fundsN0Yes“tinies”unleveredN0No
Slide37CoVaR – systemic risk measureVaRqi is implicitly defined as quantileCoVaRqj|i is the VaR conditional on
institute i (index) is in distress (at it’s VaR level)ΔCoVaR
qj|i =
CoVaRqj|i –
VaR
q
j
Various conditionings? (direction matters!)
Contribution
ΔCoVaRQ1: Which institutions contribute (in a non-causal sense)VaRsystem| institution i in distress Exposure ΔCoVaRQ2: Which institutions are most exposed if there is a systemic crisis?VaRi | system in distressNetwork ΔCoVaRVaR of institution j conditional on i
in non-causal sense!
q-prob. event
Slide38Network CoVaRconditional onorigin of arrow38
270
70
118
247
57
108
116
50
357
133116 726772122 495076
564 68
Slide39OverviewTheory (with Sannikov)Spirals: Non-linear adverse feedback loopsExternalitiesImplementation – CoVaR (with Adrian)CoVaR: measuring systemic risk contribution/externalitiesOne method: Quantile regressionsAddressing procyclicalites due to spirals39
Slide40Quantile Regressions: A RefresherOLS Regression: min sum of squared residualsPredicted value:Quantile Regression: min weighted absolute valuesPredicted value:
40
Note out (non-traditional) sign convention!
Slide41Quantile Regression: A Refresher41
Slide42Financial Intermediary DataPublicly traded financial intermediaries 1986-2008Commercial bank, security broker-dealers, insurance companies, real estate companies, etc.Weekly market equity data from CRSPQuarterly balance sheet data from COMPUSTATCDS and option data of top 10 US banks, daily 2004-200842
Slide43Change in total asset value XitChange in total asset value (detrended) where At+ = market equity * leverage ratios “detrend factor”43
Slide4444
Variable
Mean
Std. Dev.
Min
Max
Observations
Returns
overall
0.27
55.92
-2430.04
2420.38
N = 47895
between
1.07
-4.40
2.98
n = 44
within
55.91
-2431.34
2419.08
T-bar = 1088
Portfolio VaR
overall
-105.59
128.35
-1547.03
237.35
N = 47895
between
110.07
-366.58
-3.45
n = 44
within
71.16
-1433.60
493.51
T-bar = 1088
Delta CoVaR
overall
-500.76
523.62
-4956.01
2285.65
N = 47895
between
361.39
-1262.44
278.14
n = 44
within
383.75
-4488.66
2533.57
T-bar = 1088
Summary Statistics of Risk Measures OLD
Slide45ΔCoVaR vs. VaRVaR and ¢ CoVaR relationship is very weakData up to 12/0645
Slide46OverviewTheory (with Sannikov)Spirals: Non-linear adverse feedback loopsExternalitiesImplementation – CoVaR (with Adrian)CoVaR: measuring systemic risk contribution/externalitiesOne method: Quantile regressionsAddressing procyclicalites due to spiralsStep 1: Time-varying CoVaRs
Step 2: Predict CoVaR using institution characteristicsBalance sheet variables (leverage, maturity mismatch, + interdependence, …)Market variables (CDS, implied vol.,…)
46
Slide47OverviewMeasuring Systemic Risk ContributionOne Method: Quantile RegressionsCoVaR vs. VaRAddressing ProcyclicalityStep 1: Time-varying CoVaRsStep 2: Predict CoVaR using institution characteristicsBalance sheet variables
(leverage, maturity mismatch, + interdependence, …)Market variables (CDS, implied vol.,…)47
Slide48Step 1: Time-varying CoVaRControl for macro factors, Mt interpretationVIX Level “Volatility”3 month yieldRepo – 3 month Treasury “Flight to Liquidity”Moody’s BAA – 10 year Treasury “Credit indicator”10Year – 3 month Treasury “Business Cycle”Real estate index “Housing”Equity market riskObtain Panel data of CoVaR
Next step: Relate to institution specific (panel) data48
Slide49Step 1: Time-varying ΔCoVaRDerive time-varying VaRtFor institution i:For financial system:Derive time-varying CoVaRtΔCoVaRt = CoVaRt - VaR
t49
Slide50Table 2: Average Exposures to Risk Factors50
INSTITUTIONS
COEFFICIENT
VaR
system
VaR
i
CoVaR
system|i
Repo spread (lag)
-
1163***
-0.60
-877.94***
Credit spread (lag)
-107.75
-0.47
-226.75**
Term spread (lag)
128.71
0.64
18.80
VIX (lag)
-68.97***
-
0.16***
-
43.35***
3 Month Yield (lag)
118.73
0.42
15.95*
Market Return (lag)
242.74***
0.50***
196.00***
Housing (lag)
5.63
0.03
5.17
*** p< 0.01
** p<
0.05
* p< 0.1
Table 1: Summary Statistic51
Variable
Mean
Std. Dev.
Obs
X
i
overall
0.20
10.18
N = 316697
between
0.43
n
= 430
within
10.17
T-bar = 737
VaR
i
overall
-8.58
24.67
N = 316697
between
19.69
n
= 430
within
12.38
T-bar = 737
Δ
CoVaR
i
overall
-578.41
572.54
N = 316697
between
347.36
n
= 430
within
462.40
T-bar = 737
Time-varying VaR 52
Slide53Time-varying VaR and ΔCoVaR53
Slide54Step 2a: Portfolios Sorted on CharacteristicsInstitutional characteristics matter… but individual financial institutions have changed the nature of their business over timeForm decile portfolios, each quarter, according to previous quarter’s data:LeverageMaturity mismatchSizeBook-to-MarketAdd 4 industry portfoliosBanksSecurity broker-dealers
Insurance companiesReal estate companies 54
Slide55Table 3A: ΔCoVaR Forecasts by Characteristics Cross-section, Portfolios, 1%55
COEFFICIENT
2 Years
1 Year
1 Quarter
Δ
CoVaR
(lagged)
0.71***
0.80***
0.94***
VaR
(lagged)
-1.99***
-2.27***
-0.47***
Leverage (lagged)
-9.43***
-10.73***
-2.53**
Maturity mismatch
(lagged)
-0.89***
-0.30
-0.14
Relative
Size (lagged)
-170.84***
-161.99***
-38.58***
Book-to-Market
(lagged)
85.24***
87.65***
31.03**
Constant
-40.92**
-50.04**
-19.93*
Observations
3627
3805
3939
R
2
0.62
0.69
0.89
Discussion of Table 3AAt 2-year horizon, all characteristics are significantLeverage, maturity mismatch, size are positive related to systemic risk contributionHigher book-to-market indicates less systemic riskTwo effectsCloseness to default boundaryRiskiness of assetsLatter effect seems to dominate56
Slide57Table 3B: ΔCoVaR Forecasts by Characteristics Cross-section, 2 years 57
COEFFICIENT
1%
5%
10%
Δ
CoVaR
(lagged)
0.71***
0.63***
0.70***
VaR
(lagged)
-1.99***
-1.86***
-1.38***
Leverage (lagged)
-9.43***
-5.08***
-4.23**
Maturity mismatch
(lagged)
-0.89***
-0.51***
0.10
Relative
Size (lagged)
-170.84***
-105.62***
-86.84***
Book-to-Market
(lagged)
85.24***
26.95***
-14.77**
Constant
-40.92**
-14.70*
36.88***
Observations
3627
3627
3627
R
2
0.62
0.62
0.70
Discussion of Table 3BCoefficients get larger further out in the tail, indicating more $-value of assets at risk in the tailCoefficients appear significant, as beforeIn addition to including time effects as in Tables 3, we are adding fixed effects in Table 4Shows the extent to which changes to future systemic risk can be forecasted over time58
Slide59Table 4: ΔCoVaR Forecasts by Characteristics Time Series/Cross Section, Portfolios, 1%59
COEFFICIENT
2 Years
1 Year
1 Quarter
Δ
CoVaR
(lagged)
0.41***
0.58***
0.86***
VaR
(lagged)
-1.30***
-1.74***
0.06
Leverage (lagged)
0.92
-8.10***
-1.64
Maturity mismatch
(lagged)
-0.31
-0.53
-0.33
Relative
Size (lagged)
-230***
-229***
-56***
Book-to-Market
(lagged)
29.25
42.69
31.03**
Constant
-332.58***
-239.05***
-96.84***
Observations
3627
3805
3939
R
2
0.69
0.73
0.89
Timing of tail risk is harder to forecast than cross-section contribution
Slide60Step 2b: Forecasting with Market VariablesCDS spread and equity implied volatility for 10 largest US commercial and investment banks(from Bloomberg)Betas:Extract principal component from CDS spread changes/implied vol changes within each quarter from daily dataRegress each CDS spread change/ implied vol change on first principal component60
Slide61Table 6: ΔCoVaR Forecasts by Market Variables Cross Section, Portfolios, 1%61
COEFFICIENT
2 Years
1 Year
1 Quarter
Δ
CoVaR
(lagged)
0.60***
0.79***
0.94***
VaR
(lagged)
-1.84
0.05
-0.08
CDS beta (lagged)
-1.727**
787.92
95.37
CDS
(lagged)
1.320
-2.211
-40.26
Implied
Vol
beta
(lagged)
-8.30
-590.28**
-85.78
Implied
Vol
(lagged)
-144.60
111.02
234.56***
Constant
-335.30
-147.72
-114.07*
Observations
114
154
184
R
2
0.36
0.57
0.77
short data-span (2004-2008)!
Slide62Extension to our AnalysisCo-Expected Shortfall (“Co-ES”)Advantage: coherent risk measureDisadvantage: any estimate “in” the tail is very noiseInclusion of additional informationderivative positionsoff-balance sheet exposureCrowdedness measureInterdependence measuresBank supervision information62
Slide63Countercyclical RegulationWhen market is relaxedStrict Laddered ResponseStep 1: supervision enhancedStep 2: forbidden to pay out dividendsSee connection to debt-overhang problem)Step 3: No Bonus for CEOsStep 4: Recapitalization within two months + debt/equity swapWhen market is strict Relax regulatory requirement
63
Slide64Causal risk spill over effectsNon-causal64Adverse feedback loop - amplification
A
B
C
Slide65Shock Amplifier vs. Absorber OLD65INSTITUTIONS
VaR_index
VaR_index
COEFFICIENT
1 Year
1.5 Years
1 Year
1.5 Years
Fitted CoVaR_contrib (lag)
4.46**6.43***
(1.91)
(1.95)
Resid CoVaR_contrib (lag)
0.50
0.52
(0.40)
(0.41)
Fitted CoVaR_exp (lag)
0.75
0.51
(1.42)
(1.34)
Resid CoVaR_exp (lag)
2.94***
3.95***
(0.57)
(0.54)
VaR_index (lag)
0.30**
0.13
-1.25***
-1.96***
(0.12)
(0.12)
(0.33)
(0.32)
Slide66What type of charge?Capital chargeStrictly bindingMight stifle competitionPigouvian tax + government insuranceGenerates revenueIn times of crisis it is cheap to issue government debt very salientPrivate insurance scheme (Kashap, Rajan & Stein, 2008 + NYU report)Requires lots of regulation66
Slide67ConclusionTheoryLiquidity spirals - non-linear dynamicsExternalitiesMacro-prudential regulationFocus on externalitiesMeasure for systemic risk is needed, e.g. CoVaRCountercyclical regulationFind variables that predict average future CoVaR
Forward-looking measures, spreads, …Also,VaR measure is not sufficient – incorrect focusQuantile regressions are simple and efficient way to calculate CoVaR
67