Robert Engle Volatility Institute of NYU Stern New Frontiers in Systemic Risk Measures and Extreme Risk Management Brooklyn College June 4, 2015
FAILURE OF FINANCIAL INSTITUTIONS
September 15, 2008 Lehman declared bankruptcy and the most severe events of the financial crisis began.
The U.S. had arranged an orderly rescue of Fannie Mae and Freddie Mac the week before and then saved AIG, Merrill Lynch, Citigroup, Bank of America, Morgan Stanley, Goldman Sachs. Washington Mutual failed and Wachovia was purchased in the following days.
Should we rescue such firms? Should we have rescued Lehman?
If firms count on being rescued, they will take on too much risk.
Is there a third choice?
REGULATORY CHALLENGE
ENSURE THAT FINANCIAL INSTITUTIONS HAVE SUFFICIENT CAPITAL SO THAT THEY CAN NOT ONLY SURVIVE A FINANCIAL CRISIS, BUT CONTINUE TO INTERMEDIATE AND PROVIDE FINANCIAL SERVICES TO THE REAL ECONOMY.
GOOD REGULATION IS PREVENTION – NOT RESCUE
6/3/2015VOLATILITY INSTITUTE 4
SYSTEMIC RISK
WHEN THE FAILURE OF AN INSTITUTION TO MEET ITS OBLIGATIONS HAS SERIOUS CONSEQUENCES FOR THE REAL ECONOMY
FAILURE OF ONE INSTITUTION WILL BE MUCH MORE DANGEROUS IF THE WHOLE FINANCIAL SECTOR IS WEAK.
DEFINITION of SRISK
How much capital would a financial institution need to raise in order to function normally if we have another financial crisis?
We measure this econometrically based on market data on equities and balance sheet data on liabilities. We update weekly on V-LAB for US and Global financial firms. We call this SRISK. It is the basis of the Stern Systemic Risk Rankings.
Principle investigators: Viral Acharya, Matt Richardson and me at the Volatility Institute at NYU’s Stern School. Collaboration with HEC Lausanne and the Institute for Global Finance at University of New South Wales. Contributions by Christian Brownlees, Rob Capellini, Diane Perriet, Emil Siriwardane.
References: Acharya, Pedersen, Phillipon, Richardson “Measuring Systemic Risk (2010); Acharya, Engle, Richardson “Capital Shortfall, A New Approach to Ranking and Regulating Systemic Risks, AEAPP (2012), Brownlees and Engle, “Volatilities, Correlations and Tails for Systemic Risk Measurement”,2010
SRISK SRISK is computed from:
Where k is a prudential level of equity relative to assets. LRMES is the decline in equity values to be expected if
there is another financial crisis. SRISK depends upon size, leverage and risk.
( )( )( )
( ) ( )
,
, ,,1 11
i t t i
t t n t n t n t n
i t i ti t
SRISK E Capital Shortfall Crisis
E k Debt Equity Equity Crisis
LRMES ASSETSASSETS ASSEEqu
Tk k
ity EquityS
+ + + +
−
=
= + −
= − − −
SRISK IS A STRESS TEST
The stress scenario is a 40% collapse in the global equity market over six months.
The capital requirement is that, under stress, equity exceed 8% of total assets
Total Assets are measured as Quasi Assets which are accounting liabilities plus market equity
For IFRS firms, a 5.5% capital requirement is used.
DYNAMIC CONDITIONAL BETA
6/3/2015VOLATILITY INSTITUTE 9
ECONOMETRICS OF SRISK
WHAT IS BETA?
Beta measures the decline in a firm stock price as the market declines.
Beta is not constant – it is a correlation times the ratio of firm volatility to market volatility.
We use a global ETF and correct for asynchronicity in daily returns.
We estimate volatility (GJR) and correlation (DCC) models for more than 1000 global firms.
We do this every week and post it on V-LAB.
BETA FOR CITIGROUP
BETA FOR GOLDMAN
THE BASIC IDEA OF DCB
IF is a collection of k+1 random variables that are distributed as
Then
Hence:
( ), , 1,...,t ty x t T=
( ) , ,,1
, ,,
~ , , yy t yx ty ttt t t
xy t xx tx tt
H HyN H N
H Hxµ
µµ−
=
( )( )1 11 , , , , , , , ,, ~ ,t t t y t yx t xx t t x t yy t yx t xx t xy ty x N H H x H H H Hµ µ− −− + − −
1, ,t xx t xy tH Hβ −=
MLE
When all betas are DCBs, then the likelihood comes from Multivariate GARCH. Ignoring means,
And regression coefficients and residuals are computed given the estimated values of the volatility parameters.
When no betas are DCBs, then estimate regression with heteroskedasticerrors.
When some betas are DCBs, then subtract time varying coefficients and estimate constant ones.
( ) ( ) { } { }1 1max , , , , ,..., , ,...,vol fixed T TL y x y y y x x xθ
θ θ ψ β= ≡ ≡
( ) ( ) ( )2
1 , ,1 ,
' , , .5 log( )T
tt t t t t u t u t xx
t u t
uy x u V u h L h Lh
β θ θ−=
= + = = − + +
∑
( ) ( ) ( ) ( ) 1
1
, , .5 log , 'T
tt t t t
t t
yL y x H y x H
xθ ψ ψ −
=
= − +
∑
( )1, , 1 , , , , ,
ˆ ˆˆ ˆ ˆ ˆ ˆ ˆ, ' ,t xx t xy t t t t t t t yy t yx t xx t xy t w tH H w y x V w H H H H hβ β−−= = − = − ≅
IS BETA CONSTANT?
For none of these methods will beta be constant.
In the one regressor case this requires the ratio of to be constant.
This is a non-nested hypothesis because Either Or it is given by a bivariate GARCH model
, ,/yx t xx th h
, ,yx t xx th hβ=
NON-NESTED HYPOTHESES
Model Selection based on information criteria Two possible outcomes
Artificial Nesting Four possible outcomes
Testing equal closeness- Quang Vuong Three possible outcomes
ARTIFICIAL NESTING
Consider the model:
If phi is zero, the parameters are constant If theta is zero, the parameters are time
varying. If both are non-zero, the nested model may be
entertained.
( )' 't t t t ty x x vθ φ β= + +
TIME ZONES
MODEL WITH LAGS
Condition on t-2
The equation
But u can be an MA(1) and GARCH. In fact, it must have MA(1) if Ri is to be a Martingale difference.
( ),
, 2
, 1
~ 0,i t
m t t t
m t
RR N H
R−
−
, , , , , 1 ,i t i t m t i t m t i tR R R uβ γ −= + +
IMPLEMENTING THE MODEL
Assuming that the factor is priced efficiently, it will have no autocorrelation, hence the regressormatrix will be diagonal.
Then approximate
This is a good approximation because roots are very close to one for daily data.
2 2, 2 , , 2 , 2 , 1 , 1 2 , 1
, 1, , 1
, 1
/ /i t t i t m t t m t t i t m t t m t
i ti m t
m t
E R R E R E R R E R
hh
GAMMA
To compute gamma, treat as a bivariate process and apply DCC. Then lag the estimates:
These expressions do not include any variables dated t-1, as required.
The volatility approximation is the same as above but the autocorrelation is new.
, , 1,i t m tR R
2 , , 1 , 1, , 1, , 22
, 12 , 1
t i t m t i ti t i t m t
m tt m t
E R R hhE R
NESTED MODEL
Combining the constant beta and dynamic conditional beta into one regression:
Where u will be an MA(1) GARCH
( ) ( ), 1 , 2 , 3 , 4 , 1i t i t m t i t m t tR R R uφ β φ φ γ φ −= + + + +
RESULTS
In the Americas, the beta is the most important coefficient and gamma is generally insignificant.
In Europe both beta and gamma are significant but generally beta is bigger
In Asia, the most important regressor is gamma as the market is open almost a full day earlier than in the US.
BETA FOR BNP PARIBAS
Barclay’s Beta
GLOBAL SYSTEMIC RISK
WORLD SRISK – LAST 10 YEARS
6/3/2015VOLATILITY INSTITUTE 28
WHERE IS THE RISK today?
6/3/2015VOLATILITY INSTITUTE 29
AS PERCENT OF GDP
AS A PERCENTAGE OF MARKET CAP
TESTING MACROPRUDENTIAL STRESS TESTS: THE RISK OF REGULATORY RISK WEIGHTS
Viral Acharya, Robert Engle and Diane PierretJournal of Monetary Economics
SCOPE OF PAPER
Compares US and European stress tests since 2009 with V-LAB estimates at the same time.
How similar are the rankings? How well do risk weights predict risks? When something approximating the scenario
occurs, which rankings are more accurate?
SUMMARY OF RESULTS
Correlations between minimum T1 Leverage Ratio and VLAB stressed leverage ratio at the same time: CCAR12 rank correlation=.846 CCAR13 rank correlation=.877 CCAR14 correlation=.596 CCAR15 correlation=.731 EBA11 rank correlation = .570 (final leverage ratio)
Rank correlations between Risk Weights and Risk and V-LAB risk are much lower, often <0
Rank correlations between stock performance and stress ratios for EBA 2011 VLAB=.354, EBA=.046
6/3/2015 VOLATILITY INSTITUTE 35
(8 _ )*( *(1 ))CAPITAL SHORTFALL CCAR TLVR LIABILITIES MC LRMES= − + −
-20,000
0
20,000
40,000
60,000
80,000
100,000
2 4 6 8 10 12 14 16 18 20 22
CCAR_CAPSHORT SRISK
NOVEMBER 6, 2014 30 GSIBs
FSB 2014 GSIBs VLAB 2014 TOP 30 BY REGION
JPM, HSBC BARC,BNP,C,DB BAC,CS,GS,MIT,MS,RBS AGRICULTURAL BANK OF
CHINA,BANK OF CHINA,BONY,BBVA, GROUPE BPCE, CAG,ICBC, ING, MIZUHO, NORDEA, SANTANDER, SOC GEN, STANDARD CHART, STATE STREET, SUMITOMO, UBS, UNICREDIT, WELLS FARGO
BNP, CAG, DB, BARC, SOC GEN, RBS, UNICREDIT, ING, SANTANDER, COMMERZBANK, CS, NATIXIS, INTESSA, UBS
MIT, MIZ, BANK Of CHINA, SUMITOMO, AGRICULTURAL BANK OFCHINA, CCB, ICBC, BANK COMMUNICATIONS, CHINA MERCH, CHINA CITIC
JPM, C, BAC, BANCO BRAZIL, MS, GS
BASEL USES ADDITIONAL CRITERIA
Functions that cannot be easily replicated such as REPO. i.e. State Street, BONY
Not currently risky but could become risky and should be monitored. Wells Fargo, HSBC, Standard Chartered
Candidates for inclusion in the future: Commerzbank, Intessa, CCB, Bank of
Communications, China Merchant, China Citic, Banco de Brazil
These are based on Nov 2014 info. They have somewhat changed since then.
ECB AQR AND STRESS TESTS
ECB found that 25 of 130 banks examined needed capital under the stress scenario. All but 13 had already raised it.
These were headed by Monte dei Paschi diSienna and 3 other Italian banks, and included 2 in Greece, 1 in Cyprus, and one in each of several other Countries.
In VLAB these are all small and the total capital required is small.
The major banks of Europe also appear undercapitalized and this is a much greater risk.
EUROPE SINCE 2005
6/3/2015VOLATILITY INSTITUTE 44
EUROPEAN FIRMS
ASIA SINCE 2005
6/3/2015VOLATILITY INSTITUTE 46
CHINA SINCE 2005!!
DOES THIS MODEL APPLY TO CHINA?
MANY FACTORS ARE SIMILAR IN CHINA
State owned banks give new meaning to “too big to fail”.
Government guarantees distort risk taking. Regulatory incentives encourage lending that
might not be prudent. Shadow banking has uncertain governmental
protections. Sovereign has plenty of capacity to
recapitalize banks and does so when needed.
EQUITY MARKET PERFORMANCE
The Chinese equity market has gone roaring upward over the last year in spite of the slowing economy. This may be pure speculation by Chinese investors It may be a capitalization of the opening of the
Chinese capital markets such as by Hong Kong –Shanghai Connect
As a result SRISK has fallen.
Don’t ask….
Conclusions
US Financial sector is safer than it has been since the start of the financial crisis although recent weeks show deterioration.
Europe is still fragile and may be getting worse. Japan has reduced its risk but is stuck at a high
level. China SRISK has been increasing at a
breathtaking rate but is suddenly reversing itself. Possibly this is a result of new policies signaling more open capital markets.