Sei sulla pagina 1di 3

review

13

ERM

Black Swans,
Fat Tails and
Spherical Cows

here is an acceptable four letter word that is at the heart of


the current financial crisis, it is risk. Risk is at the centre
of the actuarial profession and the area of ERM (Enterprise
Risk Management) is gaining new ground in the actuarial world. But
there are many other organisations and professions that have a keen
interest in this area.
There are many things to consider that are beyond the traditional
actuarial training. Typically risk management has been the landscape
of Accountant / Risk Managers; indeed there is a whole raft of
frameworks and guidelines that do not form the core of actuarial
training. This article aims to describe where ERM fits into the
business and actuarial landscapes.

Risk Management
Risk management is a structured framework that allows the
management of risk. Risk can be broken down into many categories;
the Solvency II break down is: Market, Credit, Insurance, Investment,
Operational and Liquidity. But this is by no means a unique
dissection. In order to help bring some rigour and consistency,
ISO (International Organisation for Standardisation) has a draft of
guidelines on principles and implementation of risk management.
The basic risk framework covers a process as shown in Figure 1.
Figure 1

Strategy and Culture


The role of the culture of an organisation is often a key factor in
the level of a firms risk management competence. A key influence
on this is senior management and the importance they give to risk
management. The resilience of the firm in the light of unknown
unknowns depends on many of the softer aspects of its culture
such as Style and Shared-values, as well as the hard such as
Structure and Systems. It is not just a box ticking compliance
exercise or a sophisticated stochastic model. Improving the model
would not directly improve the business.
In order to analyse the strategic position of an entity there are
many frameworks already available, as this is not a new or actuarial
problem. Typically these frameworks begin with the vision of the
firm and then consider a wide range of internal and external
factors. There are many management frameworks (e.g. 7S,
PESTEL, 5 Forces) that assist in this process and a key part is to
think widely about the influence of the real world, and not set the
firms strategy in an analytical simplified view using a sophisticated
model. Much of the data used and thought process underlying
such a process would be qualitative. Risk management also needs
to be embedded in the way the firm operates.

A key role is played by the culture of an organisation in implementing


a framework such as this. This is particularly true for large
organisations with long histories, where the risks can emerge from
many parts of the organisation and many subcultures can exist,
making some risks hard to identify (e.g. Swiss Re, AIG).

A C T U A RY A U S T R A L I A April 2009

14

review

Once risks have been identified there are many ways to treat them
considered in the ISO guidance. The list below is not necessarily
mutually exclusive nor appropriate in all cases:




avoid
change the likelihood
change the consequences
share the risk (e.g. insure / reinsure / outsource)
retain the risk

ERM Influences
Rating agencies

The leading rating agencies (e.g. A.M. Best, S&P, Moodys and Fitch)
have been strong supporters of the development of ERM practices
and it became an important component of the financial strength
ratings process. S&P were the most active in incorporating ERM
explicitly into their ratings process as a separate rating category.

COSO (Committee of Sponsoring Organisations


of the Treadway Commission)

Once the treatment has been selected, then monitoring and review
is required to ensure that the organisations risks are developing
as expected. This is similar to the Actuarial Control Cycle (specify
a problem, develop a solution, monitor the consequences thereof,
and repeat the process) at the heart of actuarial work.

This is a US sponsored private sector initiative, formed in 1985. The


sponsors are the five main accounting associations (AICPA, AAA,
FEI, IIA and IMA). The COSO ERM framework can be visualised as
a three dimensional matrix of (i) objectives, (ii) components of risk
and (iii) by sub-entity.

Enterprise Risk Management

Sarbanes Oxley Act (SOA 2002) SOX 404

There are several definitions of ERM owned by high profile


organisations (such as the Casualty Actuarial Society (CAS),
and the Risk Insurance and Management Society (RIMS)). The
definitions share a common theme, that of bringing together the
understanding of risk from all parts of the organisation, and being
able to use this information to enhance value for the stakeholders
of the organisation.

The SOA requires that a firm undertake a Top Down Risk


Assessment (TDRA). The TDRA is a hierarchical framework that
involves applying specific risk factors to determine the scope and
evidence required in the assessment of internal control for significant
financial reporting elements.

The concept is a well established one of doing business; indeed


strategic decision making frameworks for business generally
consider a wide range of factors for the purposes of increasing
stakeholders value. A key issue is the culture of how firms go
about doing things; having a sophisticated model will not generate
sophisticated results, if fed with garbage data. Changing the model
does not change the business.

Black Swans

This is not news to actuaries, as many experienced model builders


know how sensitive results can be to assumptions that are hard
to pin down, especially in the tail of multiple loss distributions, and
are experts in interpreting the results.

A key issue of ERM is the


culture of how firms go about
doing things. Sophisticated
models will not generate
sophisticated results, if fed
with garbage data.

ACTUARY A U S T R A L I A April 2009

With all these frameworks what can go wrong?

A black swan event is a seemingly inconceivable


event that occurs infrequently and has
a massive impact. Quantitative
risk modelling, with its basis in
historic information, would not
be adequate to cover the risk of
these future events that have not
been conceived, especially when
using nice smooth mathematics
functions.
The tails and extremes of events most likely dont
follow nice smooth laws; across the range of possible outcomes
it is extremely unlikely to get a one size fits all distribution. Nice
smooth laws tend to assume rational behaviour underlying them.
The world can be discontinuous, which most mathematical models
are not, and this is where unpredictability rules and chaos reigns.
This is especially true where human behavioural factors play a
part, as considered in behavioural economics; e.g. most of us
are influenced by others (you need only look at 70s fashion for
extreme examples of this).
So, using convenient smooth mathematical models (even with
copulas) to quantify risk is like looking for your lost keys under the
nearest lamp post, as this is where the light is. Phelps (Economics
2006 Nobel Prize winner) is sceptical about too much belief in
mathematical models; seems like Fischer Black (also of Nobel

15

Prize and Long Term Capital Mangement fame) may agree with
him in hindsight! George Box also has a similar view: all models
are wrong, some models are useful.
This is not to say models arent helpful, but that they need to
be combined with experience, business acumen and judgment,
and are used properly. It is important to know what is and what
is not included. They also have a strong part to play in what
if analysis. The recurring mathematics / modelling can explain
everything claim is an ideological fiction. It also exemplifies
confirmation bias looking only for evidence that fits the thesis.
Small changes in average behaviour do not give good evidence
for whats happening in the tails, e.g. a smaller number of large
banks may increase stability on average, but due to globalisation
and interconnectedness, the tail risk becomes extreme (e.g.
Fannie May and Freddy Mac). Even a fat-tailed
distribution may understate the frequency
and size of some risks.

Fat Tails
Many loss distributions have simple
underlying assumptions in order to
make the mathematics solvable.
They provide answers, to several
decimal places and are based
on mathematics; for some people
this provides comfort.
It is likely that the process of risk
does not follow the mathematical model rules. David Viniar (CFO
of Goldman Sachs) observed in the Financial Times in 2007, that
the bank had seen 25 standard deviation moves several days
in a row. This quote was meant to point out how extreme the
market moves were, but it also points out how extremely wrong
the models were. These extreme movements are unlikely to
be from a conventional probability distribution, even a fat tailed
one. Mandelbrot suggests that multifractal techniques might be
applied to financial data to provide a better estimate for risk and
volatility. There is a risk that a model may become over simplified
and lose the link to the real world.

Spherical Cows
Spherical cow is a metaphor
for highly simplified models of
reality. It is from a mathematical
joke where the punch line is
the mathematicians solution
to improve milk production
and starts with Consider a
spherical cow. The point of
the joke is that model builders
will often reduce a problem to its
simplest form, in order to make

There is a great deal of belief in


mathematical models though
some models are useful, all
models are wrong.

calculations feasible, even though such simplification may hinder


the models application to reality. Actuaries do it with frequency and
severity (as the bumper sticker goes), and could be considered the
mathematicians of the insurance industry.
Many actuaries that build these models are well aware of the
sensitivity to assumptions, and the non-modeled risks, but this
can get lost in translation as the results promulgate to the most
senior management. Some knowledge and a view of what has not
been considered are crucial when using models to assist in decision
making. Putting some creative mathematical numbers to the risk
does not make it managed. It is similar to how a golf swing can get
wildly different results with small tweaks to input parameters, which
are fundamentally professional judgments. Many actuarial methods
involve backward looking, using past data which is out of date,
sparse and perhaps no longer relevant. Future trends may not be
predictable from the past (e.g. discontinuities / hysteresis).

What to Do?
Models have a useful part to play and clear communication of
the parts of the risk that are and are not modelled is crucial
information in order to be able to use the models as a useful input
to making strategic decisions. But too much belief in the specific
numbers could be dangerous and give false security. Models must
be connected to the real world, used and validated by a firms
management for them to become a useful tool.
Good analysis and advice from specialist advisers (who may have
wider access to data and more experience of particular areas
e.g. reinsurance) is a crucial step in an effective and efficient risk
management strategy. For example, ensuring that the reinsurance
programme purchased is robust and represents good value when
considering the risks being undertaken by the
insurance entity.
The communication skills of the actuarial
profession have never been more needed.
Jeremy Waite
waitejg@willis.com

A C T U A RY A U S T R A L I A April 2009

Potrebbero piacerti anche