Improving Risk Management

hartebeestgrassAI and Robotics

Nov 7, 2013 (4 years and 2 days ago)

60 views

Improving Risk Management

Unravelling the complexity of risk


Institute of Actuaries of Australia

ERM Seminar


20 September 2011

Neil Cantle

Joshua Corrigan

2

©

2011 Milliman

Contents

1.
Complex Systems Framework for Risk Analysis

2.
A New Toolset for Complex Risk Analysis

3.
Australian Case Study

4.
UK Actuarial Profession Risk Appetite Research

5.
Summary

Complex Systems Framework for
Risk Analysis

Section 1

4

©

2011 Milliman

Starting Point


Previous study leads us to the view that:


Risk tools need to embrace


Holism


Non
-
linearity / complexity


Human bias


Adaptation / evolution



Risk can be viewed as the unintended emergent property of a
complex adaptive system



Risks are a process and even complex risks can be spotted early

4

© 2011 The Actuarial Profession

www.actuaries.org.uk

5

©

2011 Milliman

Systems Thinking


Systems thinking is both a:



Worldview that:


Problems cannot be addressed by reduction of the system


System behaviour is about interactions and relationships


Emergent behaviour is a result of those interactions



Process or methodology to:


Understand complex system behaviour


See both the “forest
and

the trees”


Identify possible solutions and system learning


Utilise complexity science techniques for risk analysis

5

© 2011 The Actuarial Profession

www.actuaries.org.uk

6

©

2011 Milliman

Information
Theory

A New Perspective on Risk

Bayesian
networks

Psychology

Graph
theory

Complex
systems

Systems
dynamics

Behavioural
science

Cladistics

There are a lot of sciences which have
insights to offer in relation to the study
of complex adaptive systems...





...putting them together makes many
difficult risk management tasks easier,
and even possible

Cognitive
mapping

7

©

2011 Milliman

Understanding a Crisis

Symptoms

Causes

Sense
-
making

Understanding

8

©

2011 Milliman

Complex Adaptive Systems


Basic properties:


Has a purpose


Emergence


the whole has properties not held by sub components


Self Organisation


structure and hierarchy but few leverage points


Interacting feedback loops


causing highly non
-
linear behaviour


Counter
-
intuitive and non
-
intended consequences


Has tipping point or critical complexity limit before collapse


Evolves and history is important


Cause and symptom separated in time and space

Risk is the unintended emergent property of a company
(which is a complex adaptive system)

9

©

2011 Milliman

A Systems View Of Risk


Holism before reductionism (think “outcomes”)


Embrace human cognitive biases (and adjust inputs)


Admit non
-
linearity


Cope with adaptation (avoid static reporting/analyses)


Simple behaviours and feedback can produce complex outcomes


Risk is an evolutionary process not a point in time event

Complexity
-
based techniques reveal buried truths and
make the management of risk more intuitive

A New Toolset for Complex Risk
Analysis

Section 2

11

©

2011 Milliman

Cognitive Mapping
-

It’s all in your head!

Key Nodes

Key Drivers

Gaps

Source: Milliman

People form complex models in their
head of what they see/think. As your
experts describe those models it is
possible to use cognitive mapping
techniques to reconstruct the highly
complex risk profiles of real business
in a robust, repeatable way.


You can evidence areas where
narrative is too brief or where there
are conflicting views.


It is a natural way for experts to
engage but helps them combine their
thoughts with others and identify the
really important facts.

12

©

2011 Milliman

Case Study


UK Life Assurer had a series of
operational risk scenarios which were
monitored regularly and had been
modelled as loss
-
distributions


Lack of real engagement between
capital modellers and business as the
model was a bit “abstract”


Scenarios were discussed with business
experts who described the features and
dynamics of them


The scenarios were converted to a
cognitive map and analysed to elicit the
particularly key features

Cognitive map of
scenario description

...analysed to identify
key features (red)

Modelled using Decision Explorer

13

©

2011 Milliman

Case Study


A Bayesian Network
was produced from the
cognitive map for each
scenario






Business experts fine
-
tuned the model and
provided evidence to
explain the states of
each node in the model

Modelled using
AgenaRisk

14

©

2011 Milliman

Case Study


Factors which are present in
multiple scenarios are
explicitly connected


Final loss distribution
obtained by adding
scenarios together

15

©

2011 Milliman

Risk Monitoring with New Risk Metrics


Using metrics designed to describe complex non
-
linear patterns,
you can see signs of trouble building up and begin to form
theories about the dynamics



You can actually measure how much information something
contains:



I
(
x
) =
-
log
p
(
x
)



If something is surprising it will tell you a lot


Looking at your management information in this way can yield
insights about the early development of unusual behaviours

16

©

2011 Milliman

Connectivity


Typical correlation measures cannot spot
non
-
linear dependency


Mutual information sharing can

Different levels of correlation

Q

~ U[0,2
p
]

R 縠U嬴[ 㕝

堠㴠R
捯s

Q

Y = R sin
Q


Sample of 1000

Example

Correlation = 0.0

Mutual Info = 1.0

17

©

2011 Milliman

Looking beneath the surface

Produced by
Milliman using:

Same
outcome
but
different
drivers

This company’s performance seems less “complex”

This company’s performance seems complex, involving many variables

18

©

2011 Milliman

Emerging Risk


Risk registers typically force the assignment of a label to each
entry


But the entries are often not that simple


By using a more granular labelling
approach it is still possible to aggregate
the information


Technique from biology permits analysis of:


Which entries are “like” each other


Understanding of how risk scenario characteristics evolve


Clues about potential future scenarios

19

©

2011 Milliman

Evolutionary forces


Application of
Cladistics



Developed in biology to permit
classification of organisms into
groups without prejudging what
the hierarchy of relationships
should be



A simple technique gives a
much more realistic idea
about the risk profile of the
business

Source: Milliman Risk DNA Analysis


20

©

2011 Milliman

Risk Culture


Systems view of risk culture looks at


Structure of company’s communication



infrastructure (who is talking to who)


Measure efficiency of info transmission


Identify traits of company personality


key person risk


Identify current position of company’s personality from different
perspectives


Indicate current potential of company to achieve different levels from
different perspectives


Develop plan to improve maturity of risk culture
within the bounds of
what is possible


Simple questions
-
based input, but...


...scientifically grounded in psychology,
behaviourial

science, social
network analysis and complex systems

Australian Case Study

Section 3

22

©

2011 Milliman

Australian Industry Fund Case Study


Hypothetical Australian industry superannuation fund



Primary strategic objectives:


Provide retirement savings and pension products and services that
meet member needs


Maintain, enhance and protect their member value proposition



Key questions:


What are the most important drivers of the business?


How complex is the business?


How do the risks inter
-
relate and interact?

23

©

2011 Milliman

Concept Map

Industry goal

Company goal

41 concepts

81 links

24

©

2011 Milliman

What are the Drivers of the Business?

Top

10 concepts / business drivers

#

immediate
links

Weighted
links

Retain existing members

10

22

Risk and retirement

product selection

8

21

Provide attractive returns

7

19

(Poor) Capital market conditions

7

17

Ageing member

population

7

16

Maintain

low fees

6

18

Generate

economies of scale

6

19

AUM size and growth

6

18

Effective

operational and governance structures

6

16

Member contributions

6

15

25

©

2011 Milliman

Concept Map

Critical

Potent

Standard

Industry goal

Company goal

41 concepts

81 links

26

©

2011 Milliman

Most Critical Business Driver
-

Retention

27

©

2011 Milliman

Economies of Scale

28

©

2011 Milliman

Identify Feedback Loops


卣敮慲S漠瑥獴s

18 feedback
loops exist in
this business.
This is one of
them.

Use to drive
scenario tests
around concepts
not immediately
obvious

UK Actuarial Profession Risk
Appetite Research

Section 4

30

©

2011 Milliman

Risk Appetite Research


UK Actuarial Profession put out a call for research to provide
practical tools for creating a risk appetite framework and
emerging risk



Milliman and the Universities of Bath and Bristol Systems Centre
delivered a set of tools leveraging complex systems methods



It is hard to align operational risk limits to overall risk appetite as
the relationships are many and non
-
linear

31

©

2011 Milliman

Why is Risk Appetite Complex?

32

©

2011 Milliman

Risk Appetite Research

Balance
Sheet

P&L

Reputation

Credit

Market

Liquidity

Insurance

Operational

Break down high level risks into more
granular perspectives....

33

©

2011 Milliman

Risk Appetite Research

Risk appetites are linked to a series of
operational indicators whose level should
reflect the level of risk being taken

Explicit allowance for factors which relate to
multiple risks

34

©

2011 Milliman

Risk Appetite Research

Bayesian Network used to identify
what state the indicators will be in if
the risk appetite levels are reached...

35

©

2011 Milliman

Risk Appetite Research

Same model can be used to estimate
the risk level once current level of
indicators observed...

Summary and Discussion

Section 5

37

©

2011 Milliman

Summary


Studies confirm that modern society and its companies are
becoming increasingly complex


The study of complex adaptive systems brings tools to help
understand and manage such systems


Using techniques to understand “the system” makes it easier to
manage risks


Think “outcomes” not “how”


Frameworks need to be adaptive and able to cope with non
-
linearity


Don’t forget about the people

38

©

2011 Milliman

Thank You!


Questions?