Upload
others
View
1
Download
0
Embed Size (px)
Citation preview
Celta CoES : Size, interconnectedness and Black Swansin the financial system
Javier Ojea Ferreiro3
Supervised by: Alfonso Novales Cinca4
Complutense University of Madrid
May 8th, 2017
[email protected] [email protected] Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 1 / 34
Motivation
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 2 / 34
Motivation
What is and why should we care systemic risk?
• Risk of experiencing eventswhere financial failures likely totranslate into adverse effects onwelfare in the economy([ECB10]).
• Relevance in recent crisis
♦ 2008 Financial crisis: BNPParibas.
♦ 2010-2012 EuropeanSovereign Debt crisis:Monte di Paschi, AngloIrish Bank,...
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 3 / 34
Motivation
What is and why should we care systemic risk?
• Risk of experiencing eventswhere financial failures likely totranslate into adverse effects onwelfare in the economy([ECB10]).
• Relevance in recent crisis
♦ 2008 Financial crisis: BNPParibas.
♦ 2010-2012 EuropeanSovereign Debt crisis:Monte di Paschi, AngloIrish Bank,...
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 3 / 34
Motivation
What is and why should we care systemic risk?
• Risk of experiencing eventswhere financial failures likely totranslate into adverse effects onwelfare in the economy([ECB10]).
• Relevance in recent crisis
♦ 2008 Financial crisis: BNPParibas.
♦ 2010-2012 EuropeanSovereign Debt crisis:Monte di Paschi, AngloIrish Bank,...
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 3 / 34
Motivation
What is and why should we care systemic risk?
• Risk of experiencing eventswhere financial failures likely totranslate into adverse effects onwelfare in the economy([ECB10]).
• Relevance in recent crisis
♦ 2008 Financial crisis: BNPParibas.
♦ 2010-2012 EuropeanSovereign Debt crisis:Monte di Paschi, AngloIrish Bank,...
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 3 / 34
Motivation
What is and why should we care systemic risk?
• Risk of experiencing eventswhere financial failures likely totranslate into adverse effects onwelfare in the economy([ECB10]).
• Relevance in recent crisis
♦ 2008 Financial crisis: BNPParibas.
♦ 2010-2012 EuropeanSovereign Debt crisis:Monte di Paschi, AngloIrish Bank,...
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 3 / 34
Motivation
What do we need?
• Identifying SIFI s is a key step in the macro-prudential supervision([Tri09], [Con17]).
• Need of high-frequency systemic risk measures due to the accelerationof the financial turmoil phases.
• Features of SIFI s ([LRT14], [Ber10], [RW12], [Ber09], [IBF10])
♦ Too Big To Fail (TBTF )♦ Too Connected To Fail (TCTF )♦ Can complex characteristics of the probability distribution
function affect to systemic risk?
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 4 / 34
Motivation
The role of higher moments in systemic risk
• ES-based systemic risk measures overlook some stylish facts aboutfinancial returns
♦ Leptokurtic♦ Skewness♦ Joint tail dependence
• Black Swan events ([Tal07])
♦ Low probability of occurrence.♦ Huge effects.♦ ”Unpredictable”: hasard (tractable randomness, knightian risk)
and fortuit (accidental and unforeseen, knightian uncertainty).
• Ludic fallacy (model risk): confuse model with reality.
• Double bubble: underestimate probability and consequences.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 5 / 34
Motivation
What do we have so far?
• MES: mean behaviour of firm i given an aggregate shock in thefinancial system.
• CES: Absolute contribution of firm i to the returns of the financialsystem in case of crisis.
• SRISK: Capital need of firm i conditioned to a long crisis period.
• ∆ CoVaRm|i : Change in the ES of the financial system when firm isuffers a crisis.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 6 / 34
Literature review
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 7 / 34
Literature review MES
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 8 / 34
Literature review MES
Marginal Expected Shortfall
MES [AER12]
MESi ,t(α) =∂ESi ,t(α)
∂ωi ,t−1
• Measure of exposure of firms to the market in case of crisis (βi ,crisis).
• It is a market risk measure more than a systemic risk measure.([BCHP13] [BCHP17])
• Systemic and systematic risk are mixed ([GK14], [LR17], [KMSV17])
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 9 / 34
Literature review SRISK
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 10 / 34
Literature review SRISK
SRISK
SRISK [BE16]
SRISKi ,t = [Required capitalisation− Expected capitalisation ]+,
• Inputs: market and accountant data
♦ Common exposure: LRMES.♦ Size: Market capitalization.♦ Leverage: Total debt.♦ Capital requirement ratio. Europe k = 5.5% ([EJR15]).
• Accountant data
♦ Discrepancy problems.♦ Scarce and only available at a low frequency.♦ Misleading information for non-banking sectors ([SRS16]).♦ SRISK highly correlated with the leverage ratio ([SHK16]).
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 11 / 34
Literature review CES
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 12 / 34
Literature review CES
Component Expected Shortfall
CES [BD15]
CESi ,t(α) = ωi ,t−1MESi ,t(α)
• Add a size factor to the MES (TBTF ).
• ESm,t(α) =∑N
i=1 CESi ,t(α).
• Measure of contribution but not dependence. Lack of a benchmarkvalue in normal times.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 13 / 34
Literature review ∆ CoESm|i
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 14 / 34
Literature review ∆ CoESm|i
Delta Conditional Expected Shortfall
∆CoESm|i [AB16], [GE13]
∆CoESm|i ,t(β) = CoESm|i ,t(αs , β)− CoESm|i ,t(αn, β)
• Measure of contagion from the firm i to the financial system (TCTF ).
• ∆CoESi |m,t(β) measures the contagion from the financial system tofirm i .
• It is not aggregated unlike CES , SRISK .
• Size is not directly considered.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 15 / 34
Celta CoES
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 16 / 34
Celta CoES
Component Delta Conditional Expected Shortfall
r i,t
rm,t
CoVaRi |m,t(α, β)
VaRm,t(α)
MESi,t(α
)
CoES i |m
,t(α, β
)
1− β
β
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 17 / 34
Celta CoES
Component Delta Conditional Expected Shortfall
Celta CoESi ,t
Celta CoESi ,t(β) = ωi ,t−1∆CoESi |m,t(β)
• Size: TBTF (ωi ,t−1).
• Interconnectedness: TCTF (∆CoESi |m,t(α)) here .
• Considers complex distribution features.
• Obtained from the Expected Shortfall of the financial system.
• Can be aggregated as CES , SRISK .
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 17 / 34
Celta CoES
Component Delta Conditional Expected Shortfall
Relationship ESm,t(α)-Celta CoESi ,t
∆ESm,t =N∑i=1
ωi ,t−1∆CoESi |m,t(β)︸ ︷︷ ︸Celta CoESi,t(β)
+ωi ,t−1∆CoResi |m,t(β)
• ∆ESm,t : Change in the Expected Shortfall of the financial system
when a crisis arises.
• ∆CoResi |m,t(β): Ability of the institution to adapt and resist from anhazard change in the financial market.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 17 / 34
Data
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 18 / 34
Data
European database
• Source: Datastream
• Period: September 2006to the September 2016.
• Frequency: weekly basis.
• N: 201
♦ 23.88% Banks.♦ 37.31% Real Estate
oriented firms.♦ 9.45% Insurance
firms.♦ 29.36% Financial
services orientedfirms.
7%
13%
5%
15%
6%1%
19%
< 1%
5%
2%
12%
9%1%
Financial institutions gathered by country
ATBECYDEESFIFRGBGRIEITNLPT
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 19 / 34
Results
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 20 / 34
Results
Rank Correlation
Rank correlation (τKendall) with Celta CoES
Aug09 Feb150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Gaussian assumption
Aug09 Feb150
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1Student t copula and skewed t marginals assumption
SRISK MES CES ∆CoESi|m,t ∆CoESm|i,t Capitalization β
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 21 / 34
Results
Common SIFI in Top 20
Top 20 SIFIs common with Celta CoES: Student t copula with skewed t marginal assumption
Aug09 Feb150
20
40
60
80
100
%common
β & Celta CoES
Aug09 Feb150
20
40
60
80
100
%common
Capitalization & Celta CoES
Aug09 Feb150
20
40
60
80
100
%common
∆CoESm|i & Celta CoES
Aug09 Feb150
20
40
60
80
100
%common
∆CoESi|m & Celta CoES
Aug09 Feb150
20
40
60
80
100
%common
MES & Celta CoES
Aug09 Feb150
20
40
60
80
100
%common
SRISK & Celta CoES
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 22 / 34
Results
Celta CoES and CES
Nov06 Aug09 May12 Feb152
3
4
5
6
7
8
9
10x 10
−4
CES
CES & Celta CoES BANCO DE SABADELL
0
0.2
0.4
0.6
0.8
1
1.2
1.4
1.6x 10
−3
CeltaCoES
CESCelta CoES
Nov06 Aug09 May12 Feb150
0.5
1
Taildep
endence
orρ
Standard deviation, correlation & tail dependence BANCO DE SABADELL
0
50
100
Annualvolatility(%
)
τρσ
Aug09 Feb152
4
6
8
10
12
14x 10
−4
CeltaCoES
Celta CoES BANCO DE SABADELL
GaussianStudent t
Aug09 Feb152
3
4
5
6
7
8
9
10
11
12x 10
−4
CES
CES BANCO DE SABADELL
GaussianStudent t
Aug09 Feb150.02
0.03
0.04
0.05
0.06
0.07
0.08
0.09
0.1
0.11
0.12
MES
MES & ∆CoES BANCO DE SABADELL
0
0.02
0.04
0.06
0.08
0.1
0.12
0.14
0.16
0.18
0.2
∆CoES
MES
∆CoESm|i(β)
∆CoESi|m(β)
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 23 / 34
Results
Celta CoES and CES
Nov06 Aug09 May12 Feb150
0.5
1
1.5x 10
−3
CES
CES & Celta CoES MAPFRE
0
1
2
3x 10
−3
CeltaCoES
CESCelta CoES
Nov06 Aug09 May12 Feb150
0.5
1
Taildep
endence
orρ
Standard deviation, correlation & tail dependence MAPFRE
0
50
100
Annualvolatility(%
)
τρσ
Aug09 Feb150
0.5
1
1.5
2
2.5x 10
−3
CeltaCoES
Celta CoES MAPFRE
GaussianStudent t
Aug09 Feb150
0.2
0.4
0.6
0.8
1
1.2
1.4x 10
−3
CES
CES MAPFRE
GaussianStudent t
Aug09 Feb150
0.05
0.1
0.15
MES
MES & ∆CoES MAPFRE
0
0.1
0.2
0.3
∆CoES
MES∆CoESm|i(β)∆CoESi|m(β)
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 23 / 34
Results
Celta CoES and CES
Nov06 Aug09 May12 Feb150
0.005
0.01
0.015
CES
CES & Celta CoES INTESA SANPAOLO
0
0.005
0.01
0.015
CeltaCoES
CESCelta CoES
Nov06 Aug09 May12 Feb150.2
0.4
0.6
0.8
1
Taildep
endence
orρ
Standard deviation, correlation & tail dependence INTESA SANPAOLO
20
40
60
80
100
Annualvolatility(%
)
τρσ
Aug09 Feb150
0.005
0.01
0.015
CeltaCoES
Celta CoES INTESA SANPAOLO
GaussianStudent t
Aug09 Feb151
2
3
4
5
6
7
8
9
10
11x 10
−3
CES
CES INTESA SANPAOLO
GaussianStudent t
Aug09 Feb150
0.2
0.4
MES
MES & ∆CoES INTESA SANPAOLO
0
0.2
0.4
∆CoES
MES∆CoESm|i(β)∆CoESi|m(β)
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 23 / 34
Results
Aggregated risk measures
Systemic risk measures aggregated by country
Nov06 Aug09 May12 Feb150
50
100
%
SRISK as a percentage of the total
ATBECYDEESFIFRGBGRIEITNLPT
Nov06 Aug09 May12 Feb150
5
10
15x 10
4
AggregatedSRISK
SRISK in absolute values
Nov06 Aug09 May12 Feb150
50
100
%
CES as a percentage of ES of the market
Nov06 Aug09 May12 Feb150
0.05
0.1
0.15
DecompositionofES
CES in absolute values
Nov06 Aug09 May12 Feb150
50
100
%
Celta CoES as a percentage of total
Nov06 Aug09 May12 Feb150
0.1
0.2
AggregatedCeltaCoES Celta CoES in absolute values
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 24 / 34
Results
Aggregated risk measures
Systemic risk measures aggregated by sector
Nov06 Aug09 May12 Feb150
50
100%
SRISK as a percentage of the total
Bank Real Estate Insurance Financial Services
Nov06 Aug09 May12 Feb150
5
10
15x 10
4
AggregatedSRISK
SRISK in absolute values
Nov06 Aug09 May12 Feb150
50
100
%
CES as a percentage of ES of the market
Nov06 Aug09 May12 Feb150
0.05
0.1
0.15
DecompositionofES CES in absolute values
Nov06 Aug09 May12 Feb150
50
100
%
Celta CoES as a percentage of total
Nov06 Aug09 May12 Feb150
0.1
0.2
AggregatedCeltaCoES
Celta CoES in absolute values
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 24 / 34
Conclusions
Overview
1 Motivation
2 Literature reviewMESSRISKCES∆ CoESm|i
3 Celta CoES
4 Data
5 Results
6 Conclusions
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 25 / 34
Conclusions
Conclusions
• I build a market-based systemic risk measures that gathers TBTF andTCTF .
• The measure reflects stylish facts concerning financial returns, notweighted enough in the remaining measures.
• Celta CoES presents high rank correlation with size andinterconnectedness proxies. Similarities in cross-section with CES aremainly given by the size factor.
• Concerning the SRISK -Celta CoES similarities, they coincide with thefinancial crisis while Celta CoES does not undervalue non-bankingsectors.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 26 / 34
Conclusions
Thank you for your attention.Suggestions, comments, ideas,...
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 27 / 34
Conclusions
References I
T. Adrian and M. K. Brunnermeier, CoVaR, The American EconomicReview 106 (2016), no. 7, 1705–1741.
V. Acharya, R. Engle, and M. Richardson, Capital shortfall: A newApproach to Ranking and Regulating Systematic Risks, AmericanEconomic Review 102 (2012), no. 3, 59–64.
S. Benoit, G. Colletaz, C. Hurlin, and C. Perignon, A Theoretical andEmpirical Comparison of Systemic Risk Measures, Working Papershalshs-00746272, HAL, June 2013.
S. Benoit, J.-E. Colliard, C. Hurlin, and C. Perignon, Where the riskslie: A survey on systemic risk, Review of Finance 21 (2017), no. 1,109–152.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 28 / 34
Conclusions
References II
G.-D. Banulescu and E.-I. Dumitrescu, Which are the sifis? acomponent expected shortfall approach to systemic risk, Journal ofBanking & Finance 50 (2015), 575–588.
C. Brownlees and R. F. Engle, Srisk: A conditional capital shortfallmeasure of systemic risk, The Review of Financial Studies 30 (2016),no. 1, 48–79.
B. Bernanke, Financial Reform to Address Systemic Risk, 2009,Speech delivered at the Council on Foreign Relations.
B.S. Bernanke, Causes of the Recent Financial and Economic Crisis,2010, Statement of the FED Chairman before the Financial CrisisInquiry Commission.
V. Constancio, Macroprudential stress-tests and tools for thenon-bank sector, 2017, ESRB Annual Conference.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 29 / 34
Conclusions
References III
ECB, New quantitative measures of systemic risk, Tech. report,Financial Stability Review, 12 2010.
R. Engle, E. Jondeau, and M. Rockinger, Systemic risk in Europe,Review of Finance 19 (2015), no. 1, 145–190.
G. Girardi and A. T. Ergun, Systemic risk measurement: MultivariateGARCH estimation of CoVaR, Journal of Banking and Finance 37(2013), no. 8, 3169–3180.
L. Guntay and P. Kupiec, Taking the risk out of systemic riskmeasurement, Tech. report, American Enterprise Institute (AEI)Research Paper, 01 2014.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 30 / 34
Conclusions
References IV
International Monetary Fund, Bank for International Settlements, andFinancial Stability Board, Guidance to assess the systemic importanceof financial institutions, markets and instruments: Initialconsiderations: Report to the g-20 finance ministers and central bankgovernors, Staff of the International Monetary Fund and the Bank forInternational Settlements, and the Secretariat of the FinancialStability Board (2010).
J. Kleinow, F. Moreira, S. Strobl, and S. Vahamaa, Measuringsystemic risk: A comparison of alternative market-based approaches,Finance Research Letters 21 (2017), 40 – 46.
G. Loffler and P. Raupach, Pitfalls in the use of systemic riskmeasures, Journal of Financial and Quantitative Analysis (2017).
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 31 / 34
Conclusions
References V
L. Laeven, L. Ratnovski, and H. Tong, Bank Size and Systemic Risk,IMF Staff Discussion Notes 14/4, International Monetary Fund, May2014.
A. Rose and T. Wieladek, Too big to fail: some empirical evidence onthe causes and consequences of public banking interventions in theUnited Kingdom, Bank of England working papers 460, Bank ofEngland, August 2012.
C. Salleo, T. Homar, and H. Kick, Making sense of the EU wide stresstest: a comparison with the SRISK approach, Working Paper Series1920, European Central Bank, June 2016.
H. S. Scott, K. Ricci, and A. Sarfatti, SRISK as a Measure ofSystemic Risk for Insurers: Oversimplified and Inappropriate, Tech.report, Harvard Law School Working Paper, 2016.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 32 / 34
Conclusions
References VI
Nassim Nicholas Taleb, The black swan: The impact of the highlyimprobable, vol. 2, Random house, 2007.
J.-C. Trichet, Clare Distinguished Lecture in Economics and PublicPolicy, 2009, Speech at University of Cambridge organised by theClare College.
Javier Ojea Ferreiro (UCM) Celta CoES May 8th , 2017 33 / 34