Financial Network Systemic Risk Contributions - Hautsch, Schaumburg, Schienle - 14-16 December 2013

Financial Network Systemic Risk Contributions

SYstemic Risk TOmography: Signals, Measurements, Transmission Channels, and Policy Interventions

Nikolaus Hautsch - University of Vienna Julia Schaumburg - VU University Amsterdam, Tinbergen Institute Melanie Schienle - Leibniz Universität Hannover !CFE - 7th International Conference on Computational and Financial Econometrics London, December 14-16, 2013

Financial Network Systemic Risk Contributions

Nikolaus HautschUniversity of Vienna

Julia SchaumburgVU University Amsterdam, Tinbergen Institute

Melanie SchienleLeibniz Universitat Hannover

Introduction 2

Systemic Risk

Systemic risk: Breakdown risk of the financial system induced bythe interdependence of its constituents.

In crisis times, banks face

I liquidity shortage, undercapitalisation;leading to

I fire-sales, hoarding;

⇒ further enhanced by pro-cyclicality of capitalrequirements.

Consequences of interdependence:

I Spillovers of risks

I Co-movements of losses=⇒ Systemic risk network

Financial Network Systemic Risk Contributions

Introduction 3

Network-based systemic risk assessment

I Before the financial crisis 2007–2009, systemic risk was neglected byregulation authorities.

I Hellwig (2009), p. 134:“Regulatory reform must [...] address the risks generated by [...]interdependence and by the lack of transparency about systemicrisk exposure.”

I Need for a transparent measure for systemic risk that takesinterdependence (risk spillovers) into account.

Financial Network Systemic Risk Contributions

Introduction 4

Here: VaR-Based Systemic Risk Contributions

Objective: Individual banks’ contributions to system tail risk: ”stresstest”-type analysis given publicly available data

I Time-varying Value at Risk (VaR) conditional on observations of V

Pr(Xt ≤ qp,t(Xt)) = Pr(−Xt ≥ VaRp,t) = p,

where qp,t(Xt) = qp(Xt |V = Vt) is the pth cond. return quantile.

I Estimation of a time-varying reduced form relation in quantiles

VaRsq,t = gt(VaR

ip,t) = g(VaR i

p,t ,Bt)

given control variables Bt and VaR ip,t = VaR i

p(Wt).

I Time-varying systemic risk contribution

∂VaRsq,t

∂VaR ip,t

=∂gt(.)

∂VaR ip,t

=: βs|it

Financial Network Systemic Risk Contributions

Introduction 5

Our approach

Step 0 Selection of relevant tail risk drivers for each firm i :LASSO for quantiles

I other companies’ tail riskI macro environmentI individual characteristics

Step 1 Estimation of VaR i : post-LASSO quantile regression

Step 2 Measuring each i ’s time-varying contribution tosystem risk: VaRs as function of VaR i

I control variables: selected companies’ VaR js(from Step 0), macro environment

I test of significance of VaR i for VaRs

Financial Network Systemic Risk Contributions

Introduction 6

Contribution

I Identification of tail risk cross-linkages between financialcompanies (Systemic risk network)

I Estimation of and inference for two-stage quantile regressionmodel

I Identification of systemically relevant companies andquantification of systemic risk contributions

I Time-varying systemic risk rankings

Financial Network Systemic Risk Contributions

Network 7

U.S. network of risk spillovers

Financial Network Systemic Risk Contributions

Systemic risk betas 8

Systemic risk ranking during the crisis (June 08)

Rank Name βs|i2008 · 102 β

s|i2008 VaR

i2008

1 BANK OF AMERICA 2.86∗ 0.186 0.1542 AMERICAN EXPRESS 2.78∗ 0.278 0.1003 WELLS FARGO & CO 2.51∗ 0.186 0.1354 MARSHALL & ILSLEY 2.31∗ 0.516 0.0455 JP MORGAN CHASE & CO. 2.22 0.265 0.0846 PROGRESSIVE OHIO 1.97∗ 0.380 0.0527 LEGG MASON 1.96∗ 0.137 0.1438 REGIONS FINANCIAL 1.86∗ 0.107 0.1739 MARSH & MCLENNAN 1.76∗ 0.471 0.03710 STATE STREET 1.44∗ 0.171 0.08411 NY.CMTY.BANC. 1.12 0.090 0.12512 PNC FINANCIAL SVS. GP 1.09∗ 0.153 0.07113 CHUBB 1.07∗ 0.176 0.06114 TORCHMARK 1.00∗ 0.177 0.05715 CHARLES SCHWAB 0.91∗ 0.149 0.06016 CITIGROUP 0.90∗ 0.072 0.12417 MORGAN STANLEY 0.61∗ 0.074 0.08318 ZIONS BANCORP. 0.58∗ 0.058 0.10019 UNUM GROUP 0.34∗ 0.033 0.10420 UNION PACIFIC 0.27∗ 0.047 0.05621 HARTFORD FINL.SVS.GP. 0.24∗ 0.012 0.20122 FRANKLIN RESOURCES 0.17∗ 0.026 0.06423 T ROWE PRICE GP. 0.01∗ 0.001 0.102

Financial Network Systemic Risk Contributions

Case study 9

Pre-crisis case study

Excluding the financial crisis time period, can we infer companies’developments from our method?

I Repeat entire estimation and testing procedure using onlydata until June 2007

I Focus on companies that were affected by the crisis: AIG,Freddie Mac, Lehman Brothers, Merrill Lynch

⇒ Results are reasonable with respect to systemic riskiness andtail risk dependencies.

Financial Network Systemic Risk Contributions

Conclusion 10

Conclusions

I Transparent risk measure accounting for tail risk dependencies.

I Approach provides complementing qualitative (tail risknetwork) and quantitative (systemic risk ranking) informationon the U.S. financial sector.

I Although direct backtesting is not possible, plausibility checkssuggest that the method works.

I Measure may be extended in several ways.

Financial Network Systemic Risk Contributions

This project is funded by the European Union under the

7th Framework Programme (FP7-SSH/2007-2013) Grant Agreement n°320270

!!!!!!!

www.syrtoproject.eu

This document reflects only the author’s views. The European Union is not liable for any use that may be made of the information contained therein.