Decomposing systemic risk: the roles of contagion and common exposures

Abstract

Metadata

Provider:

ECB

URL:

https://www.ecb.europa.eu//pub/pdf/scpwps/ecb.wp2929~19cda4a673.en.pdf

Retrieved on:

Tuesday, April 23, 2024

Information

Tags:

Tao, CIBC, Tax, RWA, Risk, European Systemic Risk Board, Research Papers in Economics, Contagion, RT, The Big Six, NBC, International, Shock, Observation, Bank of Canada, HTC, European Economic Association, The Washington Post, Great, JPMorgan Chase, Paper, GM, Environment, Political economy, Journal of Financial Economics, COVID-19, Perception, BNS, Website, Silicon, IAT, Cifuentes, Probability, Balance sheet, RAN, Medical classification, Algorithm, Information technology, Quarterly Journal of Economics, LN, Nature, European Journal, Royal Bank of Canada, Technical report, Journal of Political Economy, Equitable Bank, Bankruptcy, RAI, PDF, Private, ECB, Policy, CHS, Supercapacitor, Social science, Journal of Financial Stability, Intelligence (journal), Elsevier, Home, Cambridge University Press, Journal, Springer Science+Business Media, Research, Classification, Regulation, News, EQB, Credit, Literature, AIK, European Central Bank, COVID, SVAR, Section 5, Management science, DRA, M4, VL, National bank, Government, ISSN, BMO, Panel, International Financial Reporting Standards, BIS, FIS, Basel III, Commerce, Scotiabank, C32, Econometric Society, Interbank, Fraud, Section 4, Bank, Schedule, VAR, Section 3, The Journal of Finance, RBC, Volcanic explosivity index, Fire, Wassily Leontief, Financial economics, Metric, Section 2, L14, Central bank, Superintendent, Bank of Montreal, Kronecker, BOC, Lithium, BCBS, Sale, Macroeconomic Dynamics, Christophe, CWB, LBC, NHA, Imperial Bank, Private equity, Quarterly Journal, National Bank of Canada, C51, Canadian Western Bank, Currency crisis, JEL classification codes, Victor Drai, L.1, MFC, Silicon Valley Bank, EB, Laurentian Bank of Canada, Federal, RA1, Series, W0, FEVD, Journal of Econometrics, Aggregate, University, FRB, MB, Financial institution, Element, Health, Book, Angels & Airwaves, Common, OSFI, GFC, Reproduction, K L, Systematic, Housing, G21, Home Capital Group, Communications satellite

Geotags:

III, DUARTE, EGG, DM, II, TD, RA, UNITED STATES, SWITZERLAND, UK, ROGERS, US, BT, GA, IL, EU, BASEL, GERMANY, EUROPE, UNITED KINGDOM, CANADA

Summary

Key Points:

Abstract
We evaluate the effects of contagion and common exposure on banks? capital through
a regression design inspired by the structural VAR literature and derived from the balance
sheet identity.
Contagion can occur through direct exposures, fire sales, and market-based
sentiment, while common exposures result from portfolio overlaps.
First, we document that contagion varies in time, with the highest levels
around the Great Financial Crisis and lowest levels during the pandemic.
Our new framework complements
traditional stress-tests focused on single institutions by providing a holistic view of systemic risk.
While existing literature presents various contagion narratives, empirical findings on
distress propagation - a precursor to defaults - remain scarce.
We decompose systemic risk into three elements: contagion, common exposures, and idiosyncratic risk, all derived from banks? balance sheet identities.
The contagion factor encompasses both sentiment- and contractual-based elements, common exposures consider systemic
aspects, while idiosyncratic risk encapsulates unique bank-specific risk sources.
Our empirical analysis of the Canadian banking system reveals the dynamic nature of contagion, with elevated levels observed during the Global Financial Crisis.
In conclusion, our model offers a comprehensive lens for policy intervention analysis and
scenario evaluations on contagion and systemic risk in banking.
This
notion of systemic risk implies two key components: first, systematic risks (e.g., risks related
to common exposures) and second, contagion (i.e., an initially idiosyncratic problem becoming
more widespread throughout the financial system) (see Caruana, 2010).
In this paper, we decompose systemic risk into three components: contagion, common exposures, and idiosyncratic risk.
First, we include contagion in three forms: sentiment-based contagion, contractual-based
contagion, and price-mediated contagion.
In this context,
portfolio overlaps create common exposures, implying that bigger overlaps make systematic
shocks more systemic.
With the COVID-19 pandemic starting
in 2020, contagion drops to all time lows, potentially related to strong fiscal and monetary
supports.
That is, our
structural model provides a framework for analyzing the impact of policy interventions and
scenarios on different levels of contagion and systemic risk in the banking system.
This provides a complementary approach to
seminal papers that took a structural approach to contagion, such as DebtRank Battiston et al.
More generally, the literature on networks and systemic risk started with Allen and Gale
(2001) and Eisenberg and Noe (2001).
The matrix is structured as follows:
1
In our model, we do not distinguish between interbank liabilities and other types of liabilities.
In other words, we can and aim to estimate different degrees
of contagion per asset class, i.e., potentially distinct parameters ?Ga .
For that, we build three major
metrics to check: average contagion, average common exposure, and average idiosyncratic risk.
N i j
et ,
Further, we define the (N ?K) common exposure matrix as Commt = [A

(20)

et ]diag (?C
?L

such that average common exposure reads,
average common exposure =

1 XX
Commik,t .
N i j
(22)

20

? c ),

The three metrics?average contagion, average common exposure, and average idiosyncratic risk?provide a comprehensive framework for understanding banking dynamics.
Figure 4 depicts the average level of risks per systemic risk channel: contagion risk, common exposure, and idiosyncratic risk.
Figure 4: Average levels of contagion (Equation (20)), common exposure (Equation (21)), and idiosyncratic risk
(Equation (22)).
The market-based contagion is the contagion due to
investors? sentiment, and the network is an estimate FEVD on volatility data.
For most of
the sample, we find that contagion had a bigger impact on the variance than common exposures.