1

#
Adaptive Stress Testing
with FNA HeavyTails
Alan Laubsch

Athos Risk,
Financial Network Analytics
&
Winhall Consulting LLC
ADAPTIVE STRESS TESTING WITH FNA HEAVY TAILS
2
PRMIA Webinar
Oct 29, 2014

Alan Laubsch
Director, VP of Risk Products

A new way to see risk with network science
3
Agenda
1. Adaptive Stress Testing Framework
2. Financial Cartography using Network Science
3. HeavyTails Systemic Risk Monitor
4
Implication of Complexity
From Predict & Control to Sense and Respond
with Dynamic Steering
5
Seek to understand hidden fault lines
While Qantas & CX rerouted, many continued business as usual:
We've own this route for many years, it's safe and that's the reason
why we are taking this route
Source: BBC
6
Adaptive Stress Testing: An ecosystem
powered by Macro and Micro intelligence
I. Macro: identify potential risks (hidden, structural)
Stress Library based on Thought Leaders
Focus on cycles (e.g., credit bubbles), amplifiers, imbalances, critical points
E.g., Robert Shiller: (a) tech bubble (2000), (b) housing bubble (2005)




II. Micro: monitor visible risk with market signals
Construct Stress Indices using traded factors to represent scenarios
Monitor market signals, focusing on outliers and critical points
Examples: vol spike in (a) tech stocks and (b) US mortgages & financials


See: Adaptive Stress Testing: Amplifying Network Intelligence by Integrating Outlier Information
(Laubsch 2014)
7
A Social Markets Hypothesis
Source: Wikipedia; see Geoffrey Moores Crossing the Chasm (1999)


1.Macro: Scenarios from Innovators
2. Micro: Market signals from Early Adopters
8
U.S. Subprime Bond Early Warning Case Study
S
p
r
e
a
d

C
h
a
n
g
e
-90.0%
-60.0%
-30.0%
0.0%
30.0%
60.0%
90.0%
120.0%
Date
7/19/06 9/5/06 10/23/06 12/8/06 1/29/07 3/16/07 5/2/07 6/19/07 8/6/07 9/21/07 11/8/07 12/28/07 2/15/08 4/4/08 5/21/08
300%+ increase in vol
from Dec 12 to 21 '06
99% VaR bands vs 2006-1 AAA spread changes
HSBC subprime disclosure triggers a 12 sd move on Feb 23
Ratings agencies initiate reviews and/or
downgrades week of July 9 '07
Source: Alan Laubsch Subprime Risk Management Lessons, RiskMetrics
2. GS exits
subprime
(Micro)
1. Robert Shiller warns of housing Bubble in 2005 (Macro)
9
Three Stages of Risk: U.S. Subprime Crisis
0
125
250
375
500
1/19/06 4/19/06 7/19/0610/18/061/19/07 4/19/07 7/19/0710/18/071/22/08 4/22/08
Dec 06: rst tremor
Feb 23 07: HSBC
subprime loss disclosed
AAA Subprime Bond Spreads (2006-1)
First ratings agency
downgrades week of
July 9 '07
bp's
1. Hidden Risk
Innovators
2. And the biggest surprise? 1. When was the biggest risk?
2. Emerging
Early
Adopters
3. Crisis


Early Majority
10
Tipping Point Dynamics require early
detection and action
Limited window of opportunity for exerting control
Hidden
Source: Building A Reputation Risk Management Capability, Diermeier & Loeb, 2011
Visionaries see it
Emerging
Early Adopters
Crisis
Early to Late Majority
11
Survey Question
How does your organization use outlier based risk signals?

1) We dont monitor outliers
2) We highlight outliers in risk reports
3) We organize meetings to discuss signicant outliers
12
Agenda
1. Adaptive Stress Testing Framework
2. Financial Cartography using Network Science
3. HeavyTails Systemic Risk Monitor
13
Network Theory: uncover hidden
patterns in complex data
Structure of links
between nodes matters

Augments, does not
replace other data
analysis methods

Network aspect is an
unexplored dimension of
ANY data
14
First Financial Maps
Fedwire Interbank Payment Network
(Fall 2001) was one of the rst
network views into any nancial
system.

Of a total of around 8000 banks, the
66 banks shown comprise 75% of
total value. Of these, 25 banks
completely connected


!"#$%: Soramakl, k. M 8ech, !. Arnold, 8.!. Class and W.L.
8eyeler, !"# !%&%'%() %* +,-#./0,1 20)3#,- 4'%56, hyslca A,
vol. 379, pp 317-333, 2007.
15
A map is a set of points, lines, and areas dened
by a coordinate system
Eratosthenes' map of the known world c. 194 BC
16
Maps reduce multidimensional data into
a two dimensional space
17
We can zoom in to details at different
levels
18
and depict key systemic features
19
Market linkages Clustering
Bilateral exposure data Asset price data Balance sheet data
Financial Cartography reveals
interconnectedness
Central nodes
#
www.fna.fi Alan Laubsch alan@fna.fi #

World's Ocean Currents
NASA Scientific Visualization Studio
21
Survey Question
Where might your organization benet from applying network
theory? [select any that apply]

1) Systemic risk monitoring
2) Counterpart credit monitoring
3) Monitoring traders / portfolio managers
4) Viewing internal data
5) Other
22
Agenda
1. Adaptive Stress Testing Framework
2. Financial Cartography using Network Science
3. HeavyTails Systemic Risk Monitor
23
Time series of asset returns
Example: Daily returns of asset
prices (e.g., ETFs)

Difcult to understand large-scale
correlation or other dependence
structures of nancial assets.

How to put the correlations and
their changes in context with
changes/returns and volatility?

Objective is to efciently
represent a complex system
moving in time
!
24
Correlation Networks
We encode correlations as
links between the correlated
nodes/assets.

Red link = negative correlation
Black link = positive correlation

Absence of link marks that
asset is not signicantly
correlated.
25
Dimensionality Reduction & Filtering
Next, we identify the Minimum
Spanning Tree (MST) and lter
out other correlations.

Rosario Mantegna (1999)
Hierarchical Structure in
Financial Markets

This shows us the backbone
correlation structure where
each asset is connected with
the asset with which its
correlation is strongest.
26
We highlight VaR outlier returns
Node color indicates last daily
return

Green = positive

Red = negative

Node size indicates magnitude
of return


27
Here be Dragons
Didier Sornette (2009)
Dragon King: Extreme events can be
predicted

Benoit B. Mandelbrot (1963)
Volatility Clustering: Large changes
tend to be followed by large changes


-> Identify VaR exceptions (return
outside 95% VaR bounds)

-> Map them as bright green or red
nodes

Track the number of outliers
each day
Highlight outliers in their context
28
HeavyTails Systemic Risk Analytics
See http://bit.ly/1j24pIH
29
Case: Escalating Systemic Risk 2014
http://bit.ly/1j24pIH
30
Tapering fears drove the largest global
daily selloffs in 2013 and 2014
31
32
First wave of escalating tremors
Energy 24 Jun The Canary: Energy Bull Phase Transition
Europe 7 Jul Europe Bull Phase Transition
Oil 9 Jul (jul 2-11)
Commodities 15 Jul
Euro 15 Jul to 4 Sep
Russia, Junk, EM 17 Jul MH 17
JPY/USD 30 Jul
US, Materials, 31 Jul

33
Canary in the coal mine: Energy
34
Classic contrarian
phase transition:
1. Low vol (complacency)
2. Peak prices
3. and then an outlier


35
Predictive stress test: Energy
36
Europe Equity Phase Transition
37
Europe Equity Phase Transition
The start of the negative systemic risk cascade
38
Junk is next
39
Junk Bonds (JNK) 10 Jul 2014
40
Geopolitical shock: MH17 Crash
See http://bit.ly/1j24pIH
41
JPY/USD Phase Tansition
42
Japanese Yen devaluation
The dollar rose 0.7 percent to 102.82 yen to cap a nine-day advance that
marked its longest winning streak against Japans currency since March
2005. Japans currency fell 0.6 percent to 137.71 per euro. Bloomberg,
Bonds Drop as Dollar Gains on Fed; S&P 500 Little Changed, 31 July 2014
43
31 July 2014
US Markets infected, equity & junk bonds
44
45
September-Oct 2014 Wave
JPY, GBP, Commod, Oil, Gold 2 Sep
AU, BRL, Asia 8 Sep
Real Estate 12 Sep
Asia 17 Sep
France, Junk 23 Sep
US Equities 25 Sep
Commod, Oil 30 Sep
Materials, Japan 1 Oct
USD (up), 3 Oct
Europe 7 Oct
Energy, JNK, JP, EU 9 Oct
Tech, JNK 10 Oct
Energy, Materials 13 Oct

46
2 Sep 2014
Next wave of outliers
47
48
Europe trades down with oil
while US stays at peak levels for a while longer
49
25 Sep 2014: US infected
Start of the most recent downside wave
50
51
17 Oct 2017 - Fed to the Rescue
the markets impressive countermoves on Friday were
widely attributed to a single factor -- reassuring words from
Federal Reserve ofcials (El-Erian)
52
Reverse Stress Testing
What conditions give rise to extreme portfolio losses?
53
9 Oct 2014
54
55
Survey Question
What risk scenarios are you most concerned about now

1) Fed Tapering
2) Equity bubble
3) Housing bubble
4) Bond bubble
5) Broad asset bubble
56
Summary: Financial Cartography
Improve risk models by representing
interconnectivity

Simplify complexity by ltering signal from noise

Key steps:
Whats a relevant map for you?
Where are you on the map?
Where do you want to go?
57



Summary & Conclusions
58
Adaptive Stress Testing: harness social
intelligence

1. Stress Library from Thought Leaders

Create StressIndex with key driving factors

Use Financial Cartography to build intuition on connected risks

2. Monitor Early Warning Signals (Outliers)

Focus on key systemic fault lines: what nodes are most likely
to be affected by cascading risk?

Apply severe predictive shocks to key nodes

59
Maps are essential for any Command Center: amplify
intelligence, build intuition, and manage systemic risk

Google Flu Trends (2013)
60
Mass collaboration platforms empower
social intelligence
!if you have an idea and I have an idea and we exchange them,
then we both have two ideas. It's nonzero. -- Dean Kamen
(Abundance by Diamandis & Kotler)
facebook graph of social relationships
61
Conclusions

An era of systemic risk: everything is connected

Develop visual intuition with Financial

Cartography

Spark an evolutionary leap in risk culture with

collaborative intelligence platforms
The future is already here. Its just not very evenly
distributed. - William Gibson
62
FNA
Alan Laubsch
Director & VP of Risk Products
alan@fna.
www.fna.
Questions?
63
64