Celta CoES: Size, interconnectedness and Black Swans in

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,...


Motivation







Motivation







Motivation







Motivation







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?


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.


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.


Literature review

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


Literature review MES

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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])


Literature review SRISK

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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]).


Literature review CES

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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.


Literature review ∆ CoESm|i

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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.


Celta CoES

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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


Celta CoES


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 .


Celta CoES


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.


Data

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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


Results

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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


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


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


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)


Results

Celta CoES and CES


0.5

1

1.5x 10

−3

CES

CES & Celta CoES MAPFRE

0

1

2

3x 10

−3

CeltaCoES

CESCelta CoES


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)


Results

Celta CoES and CES


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)


Results

Aggregated risk measures

Systemic risk measures aggregated by country


50

100

%

SRISK as a percentage of the total

ATBECYDEESFIFRGBGRIEITNLPT


5

10

15x 10

4

AggregatedSRISK

SRISK in absolute values


50

100

%

CES as a percentage of ES of the market


0.05

0.1

0.15

DecompositionofES

CES in absolute values


50

100

%

Celta CoES as a percentage of total


0.1

0.2

AggregatedCeltaCoES Celta CoES in absolute values


Results

Aggregated risk measures

Systemic risk measures aggregated by sector


50

100%

SRISK as a percentage of the total

Bank Real Estate Insurance Financial Services


5

10

15x 10

4

AggregatedSRISK

SRISK in absolute values


50

100

%

CES as a percentage of ES of the market


0.05

0.1

0.15

DecompositionofES CES in absolute values


50

100

%

Celta CoES as a percentage of total


0.1

0.2

AggregatedCeltaCoES

Celta CoES in absolute values


Conclusions

Overview

1 Motivation


3 Celta CoES

4 Data

5 Results

6 Conclusions


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.


Conclusions

Thank you for your attention.Suggestions, comments, ideas,...


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.


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.


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.


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).


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.


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.


Conclusions

∆CoESi |m,t(β)

r i,t

rm,t

r i,t

rm,tVaRm,t(αs) VaRm,t(α

−n ) VaRm,t(α

+n )

CoV

aRi|m

,t(α

s,β

)

CoV

aRi|m

,t(α

n,β

)

CoES i |m

,t(αs,

β)

CoES

i |m,t(α

n, β

)

(Minuend) (Substrahend)


Documents

Celta CoES: Size, interconnectedness and Black Swans in