[Skip to content]

Sign up for our daily newsletter
The Actuary The magazine of the Institute & Faculty of Actuaries

Integrating assets and liabilities in a consistent framework

Increased regulatory requirements, including
the forthcoming Solvency II regulations,
and increased management focus on
enterprise risk management, are leading insurance
companies to rethink their strategies for
enhancing their liability projection systems.
Currently, many insurance companies are
investigating ways to align and integrate their
approach to modelling assets and liabilities
within a robust, scalable, and auditable enterprise
risk management system that can handle
increasingly sophisticated and computationally
intensive calculations.
Historic silos
Traditionally the two sides of the insurance
business assets and liabilities have largely
operated in separate silos. Each side has had its
own methodologies, systems, and practices. For
example, in recent years many insurance companies
have enhanced, and in some cases
rewritten, their actuarial projection systems to
apply modern approaches to valuing their liabilities
on a market-consistent basis. However,
this is largely implemented in isolation of the
enterprise asset risk management systems
employed by the asset management units
within the company.
Such enterprise asset risk management systems
are typically server-based, IT-controlled
and daily-scheduled mission-critical systems
that are used by and provide ‘real-time’ reporting
to the full organisation through a ‘risk dashboard’.
Through this dashboard users can view
high-level aggregated company-wide resultsacross all areas of risk (market, credit, collateral,
and operational) and can break down into highlighted
areas of concern, determine economic
capital measures, perform ALM analyses, portfolio
optimisations and other ‘what-if’ type
These historically separate silos, areas of risk
coverage and technological deployment are
illustrated as the bottom-left and top-right
boxes in figure 1.
Future integration
Insurance companies realise that in order to allocate
and manage economic capital more effectively,
they need to bring assets and liabilities
into a similar framework, so that both sides of
the business are valued consistently and, equally
importantly, so that assets can be strategically
allocated against the company’s liabilities.
Today, many firms are starting to design and
implement the next generation of financial projection
and enterprise risk management systems
with this goal of integrating assets and liabilities
into a consistent framework. Some key
objectives for these new systems include the
ability to:
? apply a consistent framework across multiple
product lines, business units, and geographies;
? provide an interactive risk dashboard that
can be used by actuarial and non-actuarial
users across the entire organisation;
? leverage the existing liability projection calculations
and client-specific implementations;
? facilitate the modelling of increasingly
sophisticated asset instruments, and incorporate
dynamic hedging strategies within liability
capital projections;
? allow for stress testing both assets and liabilities
with a variety of scenarios across market, credit,
demographic, and operational risk factors;
? integrate with the asset management function
and allow fund managers to optimise
asset allocations and track asset returns
against the company’s liability profile;
? comply and be consistent with corporate IT
This article explores four conceptual approaches
to meeting these objectives, which are illustrated
in figure 1:
? extend liability projection systems to support market, credit, and operational risk measurement
within an enterprise risk management
? extend asset risk management systems to
include liability projections;
? implement an integrated asset and liability
risk management system that utilises liability
projection system output within an asset
enterprise risk management framework;
? implement replicated portfolio techniques
that create a proxy portfolio consisting of
standard capital market products to replicate
the scenario-dependent pay-offs generated
by the company’s existing liability projection
Under all these approaches, because the same
scenarios and consistent market-based methodologies
are used in valuing both the assets and
the liabilities, the enterprise risk system can
aggregate the results together in a reporting
1: Extend liability projection systems
One option is to enhance existing liability projection
systems by extending the asset modelling
capabilities, providing infrastructure for scalability,
and building enterprise-wide dashboard
and reporting capabilities. This approach leverages
company investment in existing liability
models but requires a significant functionality
increase to cover market, credit, and operational
risks. In addition, a significant development
effort is required to build up the enterprise level
infrastructure and reporting capabilities.
2: Extend asset risk management
At the other end of the spectrum, asset-focused
enterprise systems within the organisation
could be enhanced to model liability product
cashflows, thereby producing an integrated
asset and liability system. This approach leverages
the reporting capabilities, robustness, and
scalability of the asset-centric enterprise risk systems,
but requires significant extension to cover
liability product cashflow generation. However,
by directly projecting liability cashflows this
approach would not leverage company investment
in existing liability models.
3: Integrated asset and liability risk
management system
In contrast to the previous two approaches, this
approach seeks to maximise the functionality
and company investment in both company
liability projection and asset risk management systems by utilising liability projection pregenerated
cashflows within the asset enterprisewide
risk management system.
By projecting the liability cashflows under the
same economic scenarios as those used by the
asset risk management calculation engine, the
full enterprise-wide economic capital and ALM
profile can be measured and reported by the
enterprise risk management dashboard. However,
this approach requires the liability projection
system to project and output the full set of
liability cashflows across all required scenario
sets within an enterprise-level system.
4: Replicating portfolios
An extension of the previous approach is the
replicating portfolio approach whereby capital
market products are used to replicate the
liabilities. The idea here is to create a proxy
portfolio consisting of standard capital market
products that replicate
the scenario-dependent
pay-offs generated by
the company’s existing
liability projection systems.
As this replicating
portfolio is composed of
capital market products,
then by proxy the valuation
of the liabilities is
consistent with the valuation
of the asset side of
the balance sheet.
There are five main
steps in the process to
build a replicating portfolio:
? Use the existing liability
projection systems
to generate the liability
cashflow projections
across the defined set of
economic scenarios.
? Aggregate the liability cashflows based on criteria such as product type
or business unit. The choice of the aggregation
grouping is determined by the final analysis.
Are we allocating capital at the level of product
type or business unit? Are we hedging a particular
? Choose the universe of replicating assets. The
universe typically includes vanilla interest rate
products, market indices, and options/futures
on interest products and market indices.
? Solve an optimisation problem to determine
the portfolio of assets (from the universe of replicating
assets) that best tracks the liability cashflows
across the defined economic scenarios.
? Use the replicating portfolio as a proxy for
the liability portfolio in economic, regulatory,
hedging, or other ALM analyses.
The advantage of this approach is that the liability
projection system need only project and
output cashflows from a subset of the full
scenario set. It also more easily allows for integration
of cashflows from multiple liability projection
systems in a consistent manner and
reduces the computational requirement to run
large stochastic-on-stochastic projections. On
the other hand, it is impossible to break down
the replicating portfolios or re-aggregate the liabilities
using different grouping criteria.
Technical considerations when
implementing replicating portfolio
Traditional tools for financial markets optimise
by defining objective functions and constraints
with respect to distributional statistics standard
deviation and tracking error or, more
recently, tail measures such as value at risk or
expected shortfall. However, in the insurance
world, value distributions are unlikely to conform
to simple distributional assumptions such
as normality or even more sophisticated
assumptions that attempt to capture skewness
and kurtosis. For example, figure 2 shows the
projected cashflows in year 10 for a liability
portfolio across 500 economic scenarios.
Due to the asymmetric shape and long right
tail of the distribution, it is preferable to optimise
across scenarios. For example, consider a
proxy portfolio of assets containing treasury
bonds, swaptions, market indices, futures on
the market indices, and calls and puts on the
market indices. To create a replicating portfolio
from the proxy portfolio, we define the objective
function to minimise the sum of the norms
representing the differences between the proxy
portfolio and the liability portfolio across all 500 economic scenarios. Figure 2 also shows the
10-year future cashflows for the replicating
portfolio across the same scenarios.
Continuing this example, the mean of the
difference between the 10-year future value
cashflows of the liability portfolio and the replicating
portfolio is about 2% with a standard
deviation of 3.5%. The distribution of the cashflow
errors is more symmetric but does have a
couple of large errors in the right tail.
Another consideration is the time horizon for
matching cashflows. In the example, we
attempted to match the future 10-year cashflows.
However, there can be large errors
between the replicating portfolio and liability
portfolio in other years. Figure 3 shows how the
mean cashflow error decreases as time
approaches the target 10-year time horizon.
A better match at all dates can be obtained by
incorporating multiple tracking dates with the
same tracking attribute into the optimisation
problem. However, decreasing the error at multiple
dates may come at the expense of a marginally
higher error at the single tracking date
of 10 years.
Insurance companies are actively exploring
methods for integrating assets and liabilities in
a consistent framework for the purpose of calculating
economic and regulatory capital and
for producing other ALM reports. We have outlined
four conceptual approaches that are being
implemented to various degrees at a number of
institutions. There are trade-offs to each
approach, but all provide a platform that allows
firms to improve profitability by improving the
manner in which they manage and allocate
assets against liabilities.
Jargon buster
‘Solvency II’ is the name of the directive under
development by the European Commission that
will provide a comprehensive new framework for
insurance supervision and regulation. It is intended
to introduce across the EU a more sophisticated,
risk-based approach to supervision and capital
assessment, using modern techniques for marketbased
valuation of assets and liabilities.
Source: ABI, March 2007
Jargon buster
‘Solvency II’ is the name of the directive under
development by the European Commission that
will provide a comprehensive new framework for
insurance supervision and regulation. It is intended
to introduce across the EU a more sophisticated,
risk-based approach to supervision and capital
assessment, using modern techniques for marketbased
valuation of assets and liabilities.