Backtesting Systemic Risk Measures during Historical Bank Runs and - - PowerPoint PPT Presentation

Backtesting Systemic Risk Measures during Historical Bank Runs and the Great Depression

Christian Brownlees [UPF] Ben Chabot [Chicago Fed] Eric Ghysels [UNC] Christopher Kurz [Fed Reserve Board]

In 1907, no one had ever heard of an asset-backed security, and a single private individual could command the resources needed to bail

• ut the banking system; and yet, fundamentally, the Panic of 1907 and

the Panic of 2008 were instances of the same phenomenon, as I have discussed today. The challenge for policymakers is to identify and isolate the common factors of crises, thereby allowing us to prevent crises when possible and to respond effectively when not. Chairman Ben S. Bernanke - Speech November 8, 2013 The Crisis as a Classic Financial Panic

Systemic Risk Measurement

Systemic Risk has emerged as a key new concept in the

aftermath of the 2007–2009 Financial Crisis

Serious research efforts have been undertaken as well as the

creation of new agencies specifically designed to analyze and monitor systemic risk (e.g. the OFR in the US, the ESRB in Europe)

The number of contributions is already quite sizable. However, no

single best practice/unifying approach has clearly emerged.

Recent surveys include Bisias, Flood, Lo and Valavani (2012) and

Brunnermeier and Oehmke (2012).

Questions we try to Answer

Are systemic risk measures useful beyond standard size,

leverage indicators?

Can we predict ’bailout costs’? Is today’s banking sector more connected compared to a century

ago?

Two Papers - Work in Progress

Backtesting Systemic Risk Measures

during Historical Bank Runs and the Great Depression

Is Today’s Banking Sector more Fragile than a Century Ago?

Measuring Systemic Risk: Approaches

Many definitions have been proposed in the literature There are two main measurement approaches:

• 1. Fundamental

⇒ Stress Tests, Interbank Liquidity Networks, DebtRank,...

• 2. MarketBased

⇒ CoVaR, SRISK, Connectedness, Contagion Networks,...

Historical PerspectiveI

We take backtesting seriously and assess how useful the recently

proposed measures are when applied to historical crisis.

Ideally, one would like to look at the pre-FDIC era for a broad

enough sample of financial panics to confidently asses the robustness of systemic risk measures but pre-FDIC era balance sheet and bank stock price data were heretofore unavailable.

We rectify this data shortcoming by employing a recently

collected financial dataset spanning the 60 years before the introduction of deposit insurance.

Historical PerspectiveII

The history of banking in the US prior to WWII was fraught with

periodic financial crises and banking panics.

After the passage of the National Banking Acts of 1863 and 1864,

a national banking system was created, subject to capital requirements and regulation through the newly formed Office of the Comptroller of the Currency (OCC).

Unfortunately, the oversight and capital requirements were not

enough to provide a bulwark against a run on banks and trusts.

Historical PerspectiveIII

Start Date End Date Description Sep 1873 Dec 1873 Jay Cooke and Company bankruptcy and railroad bubble burst May 1884 Aug 1884 Brokerage firm Grant and Ward sets off banking panic Nov 1890 Mar 1890 Barings Bank crisis May 1893 Sep 1893 Bankrupcies and run on gold as an eventual result of Barings Crisis Aug 1907 Nov 1907 Failure of Knickerbocker Trust spread panic to financial trusts Jul1914 Nov 1914 Banking panic and liquidity crisis set of by WWI Aug 1921 Dec 1921 Downturn resulting from post-war monetary and fiscal contraction Oct1931 Mar 1932 Bank failures in Chicago–Britain’s Departure from gold was March 1931

Structure of Talk

The Data Systemic Risk Measures Analyzing Individual Crises Connectedness in History

The Data I

We employ financial and balance sheet data for member banks of

the New York Clearinghouse.

The NY Clearinghouse was the first clearinghouse in the US, and

it facilitated exchange, issued script, stored specie, and regulated member institutions.

Importantly, clearinghouses attempted to maintain stability of

member institutions through transparency, i.e., publishing and inspecting member balance sheets, requiring members to maintain reserves, and the provision of support in times of financial stress.

The Data II

In terms of size, the NY Clearinghouse transactions accounted for

roughly 70 % of all clearing house transactions in 1901. Member banks and trust companies of NY held deposits for most

• f the financial institutions in the US.

The NY Clearinghouse member statements are the sole source of

high-frequency balance sheet data for the time period of interest.

We will be focusing on New York Clearinghouse Banks, as we will

be able to merge the balance sheet data with financial variables necessary to estimate our measures of systemic risk.

The Data III

We collected balance statements as published by the NY

• Clearinghouse. They appeared in the Saturday morning New

York Times, Wall Street Journal and Commercial and Financial Chronicle.

We collected the data every 28 days, or 13 times a year. The data

was primarily collected from the NYT and WSJ.

The condensed balance sheets reported the average weekly and

Friday closing values of each bank’s loans, deposits, excess reserves, specie, legal tenders, circulation and clearings.

The Data IV

In some cases, missing data could not be located, as the N

Y Clearinghouse did not publish individual member information during periods of financial stress.

As pointed out by Gorton (1985), during a banking panic, the

clearinghouse organization transformed into a single firm, uniting member banks under the Clearinghouse Committee.

During some of these times the NY Clearinghouse only published

aggregate balance sheet information.

The periods for which balance sheet data was not published

include the Panic of 1873 (10/73-11/73), the Barings Crisis (12/90-2/91), the Panic of 1893 (7/93-10/93), the Panic of 1907 (11/07-1/08), and at the start of the First World War (8/14-11/14).

The Data V

The variables we collected are: capital, loans, specie (gold and

silver), circulation, deposits, legal tenders, reserves with legal depositories, and surplus.

The bank balance sheet information is supplemented with equity

data (also collected at the 28-day sampling frequency). We collect price, shares outstanding, and dividends of bank stocks trading OTC in NYC.

The Data VI

We collect data for 132 banks (112) and trusts (20) from the 6th

• f January 1866 to the 1st of December 1933. Out of these only

99 financial institutions have stock price data available (90 banks and 9 trusts).

Dropping trusts and merging to the equity returns data leaves u

s with a sample of 82 total banks. Specifically, the New York Clearinghouse published information on about 60 members in 1865, a number that slowly moves down to nearly 40 members, by the end of our sample.

Quality Control check # 1 - Deposits and Crises I

Panic of 1873 Panic of 1884 Panic of 1890 Panic of 1893

Quality Control check # 1 - Deposits and Crises II

Panic of 1907 Panic of 1914 Panic of 1921 Panic of 1931

Systemic Risk Measurement I

The systemic risk measurement literature typcally focuses on the

following objectives:

• 1. Measuring the systemic risk of individual institutions Objective:

Detect which are Systemically Important Financial Institutions (SIFI’s) that can potentially generate threats to the entire system

• 2. Measuring the systemic risk of the entire system

Objective: Produce early warnings signals that can help avoiding

• r at least mitigating a financialcrisis.

Focus here is on market based measures and both objectives.

Systemic Risk Measurement II

Backtesting Systemic Risk Measures

during Historical Bank Runs and the Great Depression

CoVaR [Adrian and Brunnermeier (2016,AER)]

tail codependence with the financial system

SRISK [Brownlees and Engle (2016,RFS)]

capital shortfall generated in times of distress

Is Today’s Banking Sector more Fragile than a Century Ago?

Connectedness [Diebold and Yilmaz (2014,JoE)]

volatility spillover effects with the rest of the financial system

Notation

ri t : compound return of banki rm t : value weighted compound return of the financial system Wi t : Market value of equity Di t : Book value ofdebt LVGi t : Leverage Ratio Di t /Wi t

CoVaR: Definition

CoVaR links the systemic risk contribution of a financial institution

with the increase of the VaR of the entire financial system which is associated with that financial entity being under stress.

CoVaR of firm i is defined as

m t it p,q q it it

P(r < CoVaR |r = VaR ) = p

q it

where VaR is the (1 ­ q)% VaR of institution i at time t.

Adrian and Brunnermeier propose to measure the systemic risk

contribution of firm i with the ∆CoVaRi t

it

∆CoVaR = CoVaR ­CoVaR

p,q p,0.50 it it

They also consider a size corrected version of the measure

∆$CoVaRit = Wit ∆CoVaRit

SRISK: Definition

SRISK links the systemic risk contribution of a financial institution

with the capital shortfall the financial institution is expected to experience in case of a substantial market downturn

The SRISK index is defined as

SRISKit = Wit[kLVGit­(1 ­k)MESit ­1] with MESit = E(ri t |rm t < C).

The parameter k is the prudential capital fraction, that is the

percentage of total assets the firm holds as reserves. We set k = 10%

The parameter C denotes the threshold loss for a systemic event.

We set C = ­5%.

In this work we also resort to a size adjusted version of the index

adj it it it

SRISK

= kLVG ­(1 ­k)MES ­1

Recall the two objectives

The systemic risk measurement literature typcally focuses on the

following objectives:

• 1. Measuring the systemic risk of individual institutions

Objective: Detect which are Systemically Important Financial Institutions (SIFI’s) that can potentially generate threats to the entire system

• 2. Measuring the systemic risk of the entire system

Objective: Produce early warnings signals that can help avoiding

• r at least mitigating a financialcrisis.

Panic of 1893 - I

Date Deposit Index CH Data Date Deposit Index CH Data 1892-06-24 1.00 yes 1893-03-03 0.85 yes 1892-07-22 0.97 yes 1893-03-31 0.81 yes 1892-08-19 0.97 yes 1893-04-28 0.80 yes 1892-09-16 0.91 yes 1893-05-26 0.81 yes 1892-10-14 0.87 yes 1893-06-23 0.77 yes 1892-11-11 0.84 yes 1893-07-21 0.71 yes 1892-12-09 0.84 yes 1893-08-18 0.70 yes 1893-01-06 0.85 yes 1893-09-15 0.70 no 1893-02-03 0.92 yes 1893-10-13 0.70 no

Panic of 1893 - II

Deposit From Growth To SRM On CoVaR SRISK Connect. Granger 1893-02-03 1893-06-23 1893-02-03

• 0.27
• 0.60
• 0.41
• 0.50

(0.039) (0.000) (0.002) (0.000) 1893-02-03 1893-10-13 1893-02-03

• 0.28
• 0.58
• 0.39
• 0.48

(0.031) (0.000) (0.002) (0.000) 1893-05-26 1893-10-13 1893-05-26

• 0.14
• 0.22
• 0.33
• 0.23

(0.295) (0.101) (0.011) (0.082) 1893-05-26 1893-10-13 1893-05-26

• 0.14
• 0.22
• 0.33
• 0.23

(0.295) (0.101) (0.011) (0.082)

Panic of 1907

Panic of 1907 - I

Date Deposit Index CH Data Date Deposit Index CH Data 1906-12-21 1.00 yes 1907-08-30 1.08 yes 1907-01-18 1.07 yes 1907-09-27 1.09 yes 1907-02-15 1.09 yes 1907-10-25 1.07 yes 1907-03-15 1.04 yes 1907-11-22 0.95 yes 1907-04-12 1.12 yes 1907-12-20 0.95 no 1907-05-10 1.13 yes 1908-01-17 0.95 no 1907-06-07 1.15 yes 1908-02-14 1.20 yes 1907-07-05 1.11 yes 1908-03-13 1.24 yes 1907-08-02 1.13 yes 1908-04-10 1.28 yes

Panic of 1907 - II

Deposit From Growth To SRM On CoVaR SRISK Connect. Granger 1907-07-05 1907-09-27 1907-07-05

• 0.07
• 0.07
• 0.09
• 0.04

(0.659) (0.646) (0.532) (0.795) 1907-07-05 1907-10-25 1907-07-05

• 0.05
• 0.10

0.08 0.07 (0.711) (0.509) (0.605) (0.623) 1907-08-02 1907-09-27 1907-08-02

• 0.28
• 0.29
• 0.25
• 0.15

(0.053) (0.045) (0.083) (0.313) 1907-08-02 1907-10-25 1907-08-02

• 0.07
• 0.13

0.01 0.08 (0.645) (0.366) (0.970) (0.584)

Panic of 1914 - I

Date Deposit Index CH Data Date Deposit Index CH Data 1914-01-16 1.00 yes 1914-09-25 0.79 no 1914-02-13 1.07 yes 1914-10-23 0.79 no 1914-03-13 1.09 yes 1914-11-20 0.79 no 1914-04-10 1.12 yes 1914-12-18 1.14 yes 1914-05-08 1.15 yes 1915-01-15 1.18 yes 1914-06-05 1.18 yes 1915-02-12 1.25 yes 1914-07-03 1.17 yes 1915-03-12 1.28 yes 1914-07-31 1.10 yes 1915-04-09 1.31 yes 1914-08-28 0.79 yes 1915-05-07 1.37 yes

Panic of 1914 - II

Deposit From Growth To SRM On CoVaR SRISK Connect. Granger 1914-05-08 1914-08-28 1914-05-08

• 0.44
• 0.55
• 0.49
• 0.44

(0.004) (0.000) (0.001) (0.004) 1914-05-08 1914-12-18 1914-05-08

• 0.32
• 0.38
• 0.34
• 0.33

(0.041) (0.015) (0.030) (0.035) 1914-05-08 1914-09-25 1914-05-08

• 0.44
• 0.55
• 0.49
• 0.44

(0.004) (0.000) (0.001) (0.004) 1914-07-31 1914-12-18 1914-07-31

• 0.04
• 0.05
• 0.04
• 0.04

(0.783) (0.759) (0.823) (0.819)

Panic of 1931 - I

Date Deposit Index CH Data Date Deposit Index CH Data 1931-02-01 1.00 yes 1931-11-01 0.86 yes 1931-03-01 1.00 yes 1931-12-01 0.82 yes 1931-04-01 0.98 yes 1932-01-01 0.79 yes 1931-05-01 0.99 yes 1932-02-01 0.76 yes 1931-06-01 0.96 yes 1932-03-01 0.73 yes 1931-07-01 0.97 yes 1932-04-01 0.78 yes 1931-08-01 0.94 yes 1932-05-01 0.79 yes 1931-09-01 0.94 yes 1932-06-01 0.76 yes 1931-10-01 0.89 yes 1932-07-01 0.77 yes

Panic of 1931 - II

Deposit From Growth To SRM On CoVaR SRISK Connect. Granger 1931-06-01 1931-12-01 1931-06-01

• 0.35
• 0.25
• 0.11
• 0.33

(0.188) (0.343) (0.689) (0.219) 1931-06-01 1932-01-01 1931-06-01

• 0.36
• 0.38
• 0.16
• 0.35

(0.165) (0.151) (0.564) (0.185) 1931-07-01 1931-12-01 1931-07-01

• 0.58
• 0.44
• 0.28
• 0.51

(0.020) (0.089) (0.288) (0.043) 1931-07-01 1932-01-01 1931-07-01

• 0.55
• 0.51
• 0.26
• 0.44

(0.029) (0.045) (0.321) (0.084)

Predicting Capital Infusion

Panic k = 10 k = 20

α0 α1

R2

α0 α1

R2 1873 0.0007∗∗∗ 0.4574∗∗∗ 0.77 0.0008∗∗∗ 0.5610∗∗∗ 0.80

(0.0001) (0.0347) (0.0002) (0.0395)

1884 0.0009∗∗∗ 0.5668∗∗∗ 0.52 0.0008∗∗∗ 0.7380∗∗∗ 0.64

(0.0002) (0.0801) (0.0003) (0.0816)

1890 0.0014∗∗∗ 0.4962∗∗∗ 0.70 0.0015∗∗∗ 0.6365∗∗∗ 0.75

(0.0002) (0.0421) (0.0003) (0.0481)

1893 0.0012∗∗∗ 0.3739∗∗∗ 0.75 0.0013∗∗∗ 0.4844∗∗∗ 0.80

(0.0002) (0.0287) (0.0002) (0.0323)

1907 0.0027∗∗∗ 0.2124∗∗∗ 0.93 0.0035∗∗∗ 0.2854∗∗∗ 0.93

(0.0006) (0.0089) (0.0007) (0.0120)

1914 0.0046∗∗∗ 0.3483∗∗∗ 0.86 0.0055∗∗∗ 0.4502∗∗∗ 0.89

(0.0013) (0.0223) (0.0014) (0.0253)

1921 0.0053∗∗ 0.3704∗∗∗ 0.89 0.0064∗∗ 0.5101∗∗∗ 0.94

(0.0024) (0.0239) (0.0025) (0.0234)

Conclusions so far

Data work is done (was torturous) - empirical analysis is very

much work in progress

Systemic measures (whatever they measure) are clearly useful -

that is the good news.

Is Today’s Banking Sector more Fragile than a Century Ago? SEQUEL COMING SOON TO A THEATRE NEAR YOU

Fourth National Bank

Fourth National 1873 Fourth National 1884 Fourth National 1914