Alarm index for institutional bank runs

• Published date: 01 July 2019
• DOI: http://doi.org/10.1002/ijfe.1715
• Date: 01 July 2019
• Author: Jan Henrik Wosnitza

RESEARCH ARTICLE

Alarm index for institutional bank runs*

Jan Henrik Wosnitza

Deutsche Bundesbank, Regional Office in

Hessia, On-Site Inspections, Taunusanlage

5, 60329 Frankfurt am Main, Germany

Correspondence

Jan Henrik Wosnitza.

Email: jan.henrik.wosnitza@bundesbank.de

Abstract

Since the insolvency of Lehman Brothers brought the global financial system

to the brink of collapse in 2008, there is again a strong interest in properly

understanding liquidity risks and the triggers of liquidity crises.

Econophysicists recently developed an alarm index for institutional bank runs

(IBRs) based on the log‐periodic power law. The key innovation of this alarm

index is that it shall measure the speculative interactions among professional

creditors that can culminate in IBRs. The paper at hand extends this line of

research, in particular, by applying new critical parameter ranges that were

recently derived directly from credit default swap (CDS) spreads of defaulted

banks. The better performance of the revised alarm index in comparison with

the originally proposed alarm index underpins the hypothesis that the CDS

market belongs to a different universality class than, for example, the stock

market. Furthermore, the refined index outperforms a modification of the bank

run probability index that was previously proposed in the financial literature.

This result further confirms the hypothesis that—under certain circum-

stances—financial markets are driven by investors whose investment decisions

critically depend on the actions of other investors.

Highlights

•We refine a recently proposed alarm index for institutional bank runs.

•Applying credit default swap (CDS) spread‐specific parameter ranges leads

to an earlier crisis detection.

•In addition, the late‐2000 financial crisis can be predicted with higher con-

fidence.

•The refined alarm index also outperforms its benchmark from the finance

literature.

•In summary, our results imply that CDS spreads belong to a new univer-

sality class.

List of abbreviations: CDS, credit default swap; IBR, institutional bank run; LCR, liquidity coverage ratio; LPPL, log‐periodic power law; PD,

probability of default; NSFR, net stable funding ratio

*

Disclaimer: The contents of this article are solely the author's responsibility. They do not necessarily reflect the view of Deutsche Bundesbank or its

staff.

Received: 21 April 2017 Revised: 3 August 2018 Accepted: 9 September 2018

DOI: 10.1002/ijfe.1715

1254 © 2018 John Wiley & Sons, Ltd. Int J Fin Econ. 2019;24:1254–1270.wileyonlinelibrary.com/journal/ijfe

KEYWORDS

bank run, CORA3 algorithm, credit default swap, econophysics, financial crisis, log‐periodic power

law

1|INTRODUCTION

“At its Peak, the financial crisis of 2007‐2009 was a crisis

of capital market funding and the drying up of liquidity

was mainly blamed on an institutional bank run”

(Ahlswede & Schildbach, 2012). From the perspective of

the paper at hand, the key term in this statement is insti-

tutional bank run (IBR). A classical run on a bank occurs

when many retail customers simultaneously withdraw

their money from a bank, because they believe that the

bank will fall into default in the near future. However,

the liabilities of a bank do often comprise not only retail

deposits but also short‐term obligations towards financial

institutions. In analogy to a classical bank run, an IBR

occurs when a high number of financial institutions

denies short‐term funding to a bank.

Before summer 2007, banks enjoyed plentiful short‐

term funding in the interbank markets. With the onset

of the financial crisis, however, the short‐term wholesale

funding market started to dry up. Initially, this dry‐up

manifested itself in higher funding costs for banks. As

the situation further deteriorated, the wholesale segments

closed, first, for individual financial institutions and,

finally, the entire short‐term wholesale funding market

froze (Gorton & Metrick, 2012). Gorton and Metrick

(2012) discover, for example, runs in the repurchase mar-

ket in the years 2007 and 2008. Complementary, Covitz,

Liang, and Suarez (2013) study runs on asset‐backed com-

mercial paper programs during the year 2007.

Many institutions could not quickly enough substitute

the lost sources of funding. Instead, they responded by

engaging in fire sales, hoarding liquidity, and curtailing

lending to the real economy. These reactions led to down-

ward spirals of asset prices and to a credit crunch (Gomes

& Khan, 2011). As a result, a severe liquidity crisis

erupted (Goldsmith‐Pinkham & Yorulmazer, 2010; Inter-

national Monetary Fund, 2011; Oet & Pavlov, 2014;

Schmieder, Hesse, Neudorfer, Puhr, & Schmitz, 2012),

which prompted governments and central banks around

the world to inject huge amounts of liquidity into the

financial system in order to stem the downward spiral

of asset prices, to support financial institutions, and, in

particular, to alleviate the impact of the financial crisis

on the real economy (Gomes & Khan, 2011).

In response to the late‐2000 financial crisis, banking

regulators introduced, among other things, the liquidity

coverage ratio (LCR).

1

This regulatory standard

prescribes European banks to hold a sufficient stock of

high‐quality liquid assets to bridge the net cash outflows

during a predefined 30‐day stress period (Van den End

& Kruidhof, 2013). Although banks have to comply with

an LCR requirement of 100% in normal times, they are

allowed to draw on their buffers and, thus, to fall below

the 100% limit in extreme situations. Van den End and

Kruidhof (2013) demonstrate that a flexible LCR require-

ment—in particular, one that tolerates less liquid assets

in the numerator of the LCR during stress—is effective

in mitigating liquidity crises.

2

However, the relaxation

of the LCR requirement as a macroprudential instrument

against liquidity crises requires regulators to have an

objective criterion for switching to the more flexible

LCR regime (Van den End & Kruidhof, 2013).

Van den End and Kruidhof (2013) propose with the

number of banks, which react to first‐round effects, and

with the share of banks, whose LCR falls below a certain

level after second‐round effects, two indicators of sys-

temic liquidity risk themselves. In the past, liquidity

shocks originated from very different sources (Gomes &

Khan, 2011; Kapadia, Drehmann, Elliott, & Sterne,

2013; Schmieder et al., 2012). The various triggers of

liquidity crises, observed in the past, suggest that the

assumption of a single best measure for liquidity risk is

illusive. In fact, a set of quantitative indicators seems to

be necessary in order to decide whether stressed condi-

tions, which could culminate in a liquidity shortfall, are

developing.

The academic literature has already some systemic

liquidity stress indicators at its disposal. For example,

Jobst (2014) combines option pricing theory with market

and balance sheet data in order to simulate the joint

probability of banks experiencing a systemic liquidity

event. Two other measures of systemic liquidity risk have

been proposed by the International Monetary Fund

(2011). The macro stress testing model applies standard

solvency stress tests at a first step and, then, assumes that

the banks are subject to liability withdrawals whose mag-

nitudes depend on the probabilities of default (PDs) of the

banks (International Monetary Fund, 2011). The basic

idea behind the systemic liquidity risk index is that a

violation of various arbitrage relationships signals a

decrease in market and funding liquidity (International

Monetary Fund, 2011). In another approach, Drehmann

and Nikolaou (2013) exploit the fact that banks submit

aggressive bids above the expected marginal rate in

WOSNITZA 1255