Unintended Systemic Risk - PowerPoint Presentation

Slide1

Unintended Systemic Risk Consequences of Regulatory Compliance

Barry

Quinn

(Queen’s University Belfast)

Barbara

Casu

(Cass Business School)

Sami Ben

Nacuer

(IMF)

Rym

Ayadi

(HEC Montreal)Slide2

Today

Motivation

Contribution

Key Finding

Our Empirical

Design

Key Findings and Discussion

Concluding remarksSlide3

MotivationIn 2012 a comprehensive review of the Basel Committee on Banking Supervision’s Core principles for effective banking supervision(BCPs)sought

to:

More effectively

deal with Systemically Important Banks

Focus on systemic risk by applying

a

macro

perspective to the

microprudential

supervision.

F

ocus

on

reducing

both the probability and impact of a bank failure.

Horvath and Wagner (2017) argue that the systemic effect of countercyclical capital requirements is ambiguous.

Danielsson

(2002) critiqued Basel II capital policies as suffering from the fallacy of

composition.

R

isk

is endogenous to the system while nonetheless appearing exogenous to individual firm

.Slide4

ContributionWe contribute to the on-going policy debate by assessing whether international regulatory compliance predicts

systemic risk

.

We focus specifically on the adoption of international capital standards and the Basel Core Principles for Effective Bank Supervision (BCP

).

We further disaggregate the systemic risk consequence of regulatory compliance into:

Supervisors actions.

Banks actions.

We

evaluate how compliance with BCP

predicts

systemic risk of

993

publicly listed banks from North America and Europe over the period

1999-2015.Slide5

Key FindingsWe find a positive predictive

relationship between BCP compliance and systemic risk.

Encouragingly

, what supervisors are doing reduces systemic risk.

Surprisingly, what supervisors want banks to do increases systemic risk.

The latter findings is driven by capital adequacy compliance.Slide6

Our Empirical DesignWe

create annual aggregate % compliance scores and disaggregated scores for banks and supervisors actions for 25 country over the period

1999-2015

.

Systemic risk is measured using a time-varying version of

Conditional Value at Risk

introduced by Adrian and

Brunnermeier

(2016)

The precedent sensitivity

of systemic risk to BCP compliance is assessed using a dynamic Difference in Difference framework.

 Slide7

Our Empirical DesignThree systems are defined to estimate systemic risk and empirical test the above relationship

A system of systemically important banks

A system of North American banks

A system of European Banks

Systemic

risk is estimate weekly using all publicly listed financial institutions in each system for the longest possible period that reliable data is available; 1990-2016.

Systemic risk is then aggregated to the quarterly level and matched with firm characteristics and BCP

scores

.

To control of the changing nature of our unbalanced panel over the period we use a portfolio

sort as per Adrian and

Brunnermeier (2008) Slide8

SampleDaily equity price data,

and quarterly balance sheet data for period

1993-2015 are sourced from Bloomberg.

The macroeconomic state variables are collected from Thomson Reuters

Datastream

.

The principal variable of interest, BCP compliance, is derived from the IMF and World Bank Basel Core Financial Sector Assessment Program (FSAP) database.

After

a number of sanity filters we have an unbalanced panel of

993

banks over a 1719 week period

Importantly this includes the publicly quoted global and domestic systemically important financial institutions identified by the FSB and EBA in the last three years.Slide9

Summary Statistics

Systemically Important Banks

European Banks

North American Banks

Variables

Mean

Std Dev

Obs

Mean

Std Dev

ObsMean

Std Dev

Obs

Δ

$

CoVaR

i95,t1173.852250.784378264.291276.097828137.14789.9339509 Δ$CoVaRi99,t2320.444542.144378370.181695.827828290.921696.4239509 VaRi95,t45.5224.12449648.3921.57827326.814.840430 VaRi99,t80.3437.31449676.6225.56827359.5429.4740430 Leverage %92.763.1544967.583.77827310.457.4140430 Size1.100.1944870.830.2181840.630.2440194 Maturity Mismatch %13.6611.4044960.130.1082730.050.0740430 Boom0.874.2944961.023.6882730.764.0140430 Overall BCP %12.597.7431818.979.418795.260.022257 Supervisors BCP %15.308.5131820.9210.968796.080.622257 Banks BCP %17.927.8531819.811.487918.592.642257 Capital Adequacy BCP %23.147.2931824.028.02879200.12257 Credit Risk BCP %23.087.2331827.529.70879210.12257 Concentration Risk BCP %31.269.9431830.7910.3587939.503.122257 Market Risk BCP %25.8511.1131829.3512.0387921.509.372257 Liquidity Risk BCP %29.129.9831826.179.6487938.585.132257 Operational Risk BCP %23.909.6631831.1713.94879190.22257 

Note: The table reports statistics of the quarterly variables used in the

Δ$CoVaR

used in the dynamic

DiD

regressions. The data spans 1990:Q1-2015:Q4 and covers 993 banks, 209 in Europe and 784 in North America.

VaR

i

q,t

is expressed in quarterly percent.

Δ$CoVaR

i

q,t

is

normalised

by the cross sectional average of market equity each quarter and is expressed in quarterly basis points.Slide10

Systemically Important Banks

Dependent variable:

Panel A: Δ

$

CoVaR

i

95,t

Panel B: Δ

$

CoVaRi99,tOverall BCP8.31*** 9.93***  (1.74)

(4.04)

Supervisors BCP

-1.22***

 -7.36***  (0.47) (1.09)Banks BCP 6.71*** 8.41***  (2.18) (1.26)Other coefficients estimates redacted for exposition purposesFixed EffectsYesYesYesYesObservations3,0823,0823,0823,082R20.610.610.600.60Adjusted R20.610.610.600.60This table reports coefficients from difference in difference regression for Δ$CoVaRi95,t on one year lag of the BCP variables, state variables, and bank characteristics in panel A and for Δ$CoVaRi99,t in panel B. Each regression has a panel of firms. Newey-west standard errors allowing for up to five periods of autocorrelation are displayed in parentheses *p<0.1;**p<0.05;***p<0.01. Bank characteristic and state variable coefficients have been redacted for visual exposition.199 (20*9.93) bps of quarterly market equity losses at the original BCP scaleSlide11

North American Banks

Dependent variable:

Panel A: Δ

$

CoVaR

i

95,t

Panel B: Δ

$

CoVaRi99,tOverall BCP19.04** 31.28**  (8.63)

(15.81)

Supervisors BCP

-45.28**

 -72.26***  (6.24) (19.73)Banks BCP 53.10*** 131.15***  (20.18) (49.85)Other coefficients estimates redacted for exposition purposesFixed EffectsYesYesYesYesObservations34,87634,87634,87634,876R20.250.250.240.24Adjusted R20.250.250.240.24Note: This table reports coefficients from difference in difference regression for Δ$CoVaRi95,t on one year lag of the BCP variables, state variables, and bank characteristics in panel A and for Δ$CoVaRi99,t in panel B. Each regression has a panel of firms. Newey-west standard errors allowing for up to five periods of autocorrelation are displayed in parentheses *p<0.1;**p<0.05;***p<0.01.626 (20*31.28) bps of quarterly market equity losses at the original BCP scaleSlide12

European Banks

Dependent variable:

Panel A: Δ

$

CoVaR

i

95,t

Panel B: Δ

$

CoVaRi99,tOverall BCP1.32*** 1.97***  (0.25) 

(.70)

Supervisors BCP

-0.62***

-0.40***  (0.20) (0.17)Banks BCP 1.89*** 2.15***  (0.58) (1.01)Other coefficients estimates redacted for exposition purposesFixed EffectsYesYesYesYesObservations6,7946,7946,7946,794R20.510.510.540.54Adjusted R20.510.510.530.53Note: This table reports coefficients from difference in difference regression for Δ$CoVaRi95,t on one year lag of the BCP variables, state variables, and bank characteristics in panel A and for Δ$CoVaRi99,t in panel B. Each regression has a panel of firms. Newey-west standard errors allowing for up to five periods of autocorrelation are displayed in parentheses *p<0.1;**p<0.05;***p<0.01.39 (20*1.97) bps of quarterly market equity losses at original BCP scaleSlide13

Bank BCP principle level Analysis

Panel A: SIBs

Panel B: North American Banks

Panel C: European Banks

Dependent variable:

Δ

$

CoVaR

i

95,

Δ$CoVaRi99,Δ$CoVaRi95,Δ$CoVaRi99,Δ$CoVaRi95,Δ$CoVaRi99,

Supervisors BCP

-23.30**

-44.49**

-21.74***

-27.99***

-13.38***

-16.68*** (9.02)(6.95)(9.92)(9.21)(4.46)(6.37)Capital Adequacy 130.35**141.33***77.82***82.60***92.63***94.20*** (8.56)(17.18)(12.32)(30.99)(3.37)(4.68)Credit Risk-11.02-15.38-9.02**-8.38-11.24***-12.20** (15.04)(31.18)(5.04)(5.18)(3.87)(5.39)Concentration Risk-2.04-5.44-6.58-9.37-5.30-7.80 (10.43)(20.88)(13.74)(27.66)(4.28)(6.28)Market Risk-25.14**-44.31***-15.14**-24.21***-30.03**-20.50*** (11.52)(3.78)(3.52)(3.78)(3.65)(5.59)Liquidity Risk-33.14***-30.28***-23.10***-20.11***-30.74***-36.97*** (14.50)(6.04)(7.10)(4.04)(6.12)(9.11)Operational Risk-11.55-12.85-12.15-9.15**-11.39-8.41 (13.63)(29.45)(9.13)(5.45)(3.69)(5.44)Observations3,0823,08234,87634,8766,7946,794R20.610.600.250.240.510.54Adjusted R20.610.600.250.240.510.53 Note: This table reports coefficients from difference in difference regression for Δ$CoVaRi95,t on one year lag of the BCP variables, state variables, and bank characteristics in panel A and for Δ$CoVaRi99,t in panel B. Each regression has a panel of firms. Newey-west standard errors allowing for up to five periods of autocorrelation are displayed in parentheses *p<0.1;**p<0.05;***p<0.01.1640 (20*82.60) bps of quarterly market equity losses at original BCP scaleSlide14

Concluding RemarksOverall there is evidence that compliance has

positive predictive

effect on systemic risk

.

This effect is most prominent in the North American system equating to 6.26% of quarterly market equity return losses.

Encouragingly what supervisors are doing reduces systemic risk

Banks’ BCP compliance increases systemic

risk and is driving by capital adequacy policies

This latter finding

suggests

that capital regulation based on a bank’s own risk can have the unintended consequence of extenuating systemic

risk (Acharya ,2009)Slide15
Slide16

Dynamic DiD Regression

BCP assessment are applied to different countries at different points in time.

Multiple groups and time periods allows us to examine the effects of BCP compliance by applying a general framework

DiD

introduced by Bertrand et al (2004).

Furthermore we use a forward looking specification to better understand the (countercyclical) macro prudential policy implications.

This specification captures the stylised fact that systemic risk is building up in the background, especially during low volatility periods.

Where

i

indexes individual banks, j indexes countries, sys indexes system, and t indexes quarters.

group/time period covariate (

)

Bank specific controls for systemic risk drivers (leverage, size, and maturity mismatch and Boom indicator)

captures the sensitivity of bank systemic risk to BCP compliance.

 Slide17

Systemic RiskSystemic risk has two economic components:Volatility paradox

: a build up of background risk in times of credit boom when contemporaneous risk is low (time series component)

Spillover

effects:

the amplification of adverse shocks in times of crisis (cross sectional component)

Contemporaneous measures of systemic risk will only capture 2

Regulation that relies on contemporaneous measures suffers from a

procyclicality

pitfall.

Unnecessarily tight after period of adverse shock.

Unnecessarily loose in period of financial stability.