assumptions of classical analysis have often failed. assumptions of classical analysis have often failed.
performance from existing risk managemen t and other opera-
tional tools by showing how best they can be applied. C o ns ide r
include premiums, liabilities on individual businesses or as-
sets, expenses, commissions, or any data publicly available
under regulatory review. By using data that are more inter-
nal to the organization, risk managers can better identify the
source of complexity and, hence, the source of risk. Opera-
tional performance, work flow statistics, sales by region, and
other more detailed business-operations data can help focus
tells a compelling story that gets to the source of risk. And
it does this with an audit trail and with robust, replicable
results that can make regulators, ratings agencies, executives, and shareholders confident that strategic objectives
are being pursued prudently.
If performance isn’t making sense, looking at these headline
metrics in isolation, without considering the complexity underlying them, can give false confidence. The more rigid, complex,
or uncertain a system like a corporation becomes, the more
fragile are its operations. Measuring both the complexity and
uncertainty enables us to target the potential warning signs of
this fragility. Using measures based on these concepts can help
raise alarm bells some time before the headline metrics indicate
trouble and provide valuable information about where to place
resources in managing risk.
NEIL J. CANTLE, an associate of the Society of Actuaries
and a fellow of the Institute of Actuaries, is a principal and
consulting actuary with the london office of Milliman. He
can be reached at firstname.lastname@example.org. BRADLEY
M. SMI TH, a fellow of the Society of Actuaries and a
member of the Academy, is chairman of Milliman. He can be
reached at email@example.com.
OECD 2009, “Innovation in Country Risk Management,”
Swiss Re 2009, “Scenario Analysis in Insurance,”
Complexity analysis can be used hand in hand with other
tools. If there are 10 variables in your analysis and nine of them
appear normal but one is dominating the signature, you can look
more closely at that business function by collecting more variables
from that area or by conducting more focused psychological assessments and applying more detailed cognitive mapping to root
out the source of risk. Such an approach also brings out the best
This article is solely the opinion of its authors. It does not express
the official policy of the American Academy of Actuaries; nor does it
necessarily reflect the opinions of the Academy’s individual officers,
members, or staff.
As the risks that were missed by traditional tools keep mounting, it seems easy to lose faith in traditional methods. Considering the vast amount of resources and regulatory inertia behind
the standard practices, it may also be tempting to simply push
forth with fancier stochastic models that make the same fundamental mistakes. What’s needed instead is a paradigm shift
in the conceptual framework.
Behavioral risk management doesn’t have to replace all the
existing tools and lessons from enterprise risk management. One
major problem with traditional stochastic models is that they’re
based on approximations that don’t match up to changing conditions and are structured to replicate behaviors observed at a
point in time. The outputs from tools like complexity analysis can
create new parameters for existing stochastic models with more
realistic assumptions in an iterative, continually updating process
and also ensure that the behaviors being modeled are those currently being displayed. These same outputs also help identify the
shortcomings of current tools, which can inform the creation of a
new model. Ultimately, this is a framework for viewing risk that
puts all existing tools in context and highlights gaps.
We’re taking forward thinking to the next level.
Find out how at: www.jpmorgan.com/cbs/actuaries
Behavioral risk management is appealing because, with
all the scientific rigor that goes into its toolbox, the output