And Now, Spread-at-Risk
Gabriel Bousbib, head of risk management at Reuters America, explains why Spread-at-Risk is a better risk measure for non-leveraged financial
The simplicity of the Value-at-Risk concept has led many to recommend
that it become a standard risk measure, not only for financial institutions
involved in large-scale trading operations, but also for retail banks, insurance
companies, institutional investors and nonfinancial concerns. The hope is
that VAR could be used by investors and analysts to compare one firm's equity
capital with that of another, the same way more traditional financial ratios
To be relevant in risk management, however, VAR depends on two key elements-the ability or need to mark-to-market the firm's assets and liabilities, whether
they be on or off the balance sheet, and the need for the firm to meet short-term
large negative variations in the market value of its assets through its
equity capital. One or both of these elements cannot easily be applied to
firms such as insurance companies, retail banks, institutional investors
and nonfinancial corporations.
The reluctance shown by these firms to adopt FASB 107 and FASB 115, which recommend the adoption of a mark-to-market approach for all assets and liabilities,
goes beyond the issue of the volatility induced by marking-to-market assets
and liabilities. It also reflects their belief that mark-to-market does
not provide an accurate economic picture of their companies' financial status.
Let us take the case of a typical retail bank in the United States whose simplified balance sheet could look as follows:
With the probable exception of its mortgage loan portfolio, it will be
impossible for this retail bank to mark-to-market its assets and liabilities
accurately. For example, one could model demand deposits as a zero coupon
perpetual debt. In practice, however, the behavior of demand deposits will
depend on alternative deposit products offered by the marketplace, together
with the necessity of some businesses and individuals to keep demand deposits.
Similarly, credit card receivables could be analyzed as a series of fixed or floating cash flows, taking into account the various caps and floors
imbedded in the credit card (for example, maximum or minimum rate on a floating
rate card, option to switch from a fixed-rate card to a floating-rate card
and vice-versa). In practice, however, the behavior of credit card receivables
depends not only on the current level of interest rates, but also on exogenous
factors such as customer convenience; seasonality; the bank's pricing and
fee policy; the bank's ability to modify the terms of the transaction by
modifying, for example, the maximum rate charged on a credit card; and so
on. Similar conclusions could be drawn for insurance products.
Contrary to the Treasury world, in which the value of a transaction is
based solely on the current level of the relevant market variables, a retail
bank or an insurance company must then incorporate additional dimensions,
including customer behavior, pricing policies and so on. These differences
are illustrated in the table below.
Marking-to-market doesn't provide actionable information to a retail
bank or the insurance company's management. The need for marking-to-market
transactions in the wholesale and treasury worlds is quite obvious. The
organization depends mostly on "bought" money, which can be shifted
very quickly. The organization's leverage via this professionally managed
"bought" is significant and the organization must thus ensure
on a daily basis, or even intraday, that it has sufficient equity capital
to meet unexpected losses, leading to the concept of Value- at-Risk.
But at a retail bank or an insurance company, the source of funds is
much more stable and "stickier," originating from a large number
of small depositors. The organization's objective is to maximize its net
margin, that is, the yield generated by its assets over the cost of its
liabilities, over a fairly long time period (such as one quarter or one
year). The potential variations of the organization's net margin as a function
of market variables, as well as exogenous factors discussed earlier, is
the relevant actionable information that must be provided to management.
Daily mark-to-market and VAR analyses do not provide any indication of
these variations. In the Treasury world, market risk management should aim
at measuring and managing the "instantaneous" (that is, daily)
changes in mark-to-market. But market risk for a retail bank or an insurance
company should focus on measuring and managing the volatility of the firm's
net returns (asset return minus liability yield) over a given time horizon
(such as monthly or annual).
To its credit, VAR, as currently defined, does a good job of addressing
the needs of a highly leveraged institution interested in potential losses
resulting from large market movements. The VAR measure provides debt and
equity holders with a measure of the equity capital required to sustain
market losses for which the market may not be willing to provide short-term
funding. But this measure is clearly of limited value to the stakeholders
of a non-leveraged financial institution.
A better and more actionable statistic for the non-leveraged financial
institution could be developed around the concept of "Spread-at-Risk,"
which would measure an institution's anticipated net spread (assets minus
liabilities) and its expected distribution over a reporting period (quarter,
semester, year). The Spread-at-Risk would need to include the institution's
projected pricing policies and incorporate the behavior of its clients.
In addition to presenting senior management with an integrated risk profile
of the institution, such a measure can provide a non-leveraged institution
with multiple benefits, including improved pricing policies, better product
design and optimized client targeting.
Let us show how Spread-At-Risk would be implemented for a retail bank.
As interest rates vary and the bank's pricing levels relative to the market
rates are modified, the yield or the cost for each bank product (expressed
either in percentage rate or in dollar amounts) could vary for a number
Market movements. Rates could decrease, thus decreasing the cost
of a CD portfolio by a factor proportional to the change in market rates.
Product "contractual features." Rates could rise and
a cap imbedded in revolving credit lines could go in the money, thus capping
the yield of the asset at the cap rate.
Retail bank pricing policies. Rates could rise and prices on super saving accounts might not be adjusted accordingly, thus triggering a wave
of withdrawals; the balances lost must then be funded at the current market
Customer behavior. During the holiday season, retail card balances are likely to increase, thus increasing the dollar amounts earned by the
retail card business.
Clearly, in practice, a retail bank may have hundreds of "products," and the concept of product in this discussion might represent groupings
of products directly available to the bank's customers. The level of product
aggregation would depend on a number of practical considerations, including
data availability, modeling complexities and the level of granularity required.
For each product, over a given time horizon (say a month or a quarter)
and for given market scenarios (generated via a historical simulation, a
covariance or Monte Carlo analysis), a modeling function would then determine
two sets of data. The first would be the expected yield or cost of the product
("Yield Matrix"), and the second would be the expected revenue/cost
of the product, after taking into account the outstanding dollar balance
of the product ("Dollar Matrix").
This information would be determined based on interest rate levels; pricing relative to the market; and forecasted balances based on expected new business,
redemptions, renewal rates and so on.
Contrary to a traditional Value-at- Risk analysis, the Spread-at-Risk
in the case of a retail bank must take into account pricing policies relative
to the market, which in turn can impact the outstanding balances on given
Each box in the first table represents the expected yield or cost of
the product for a combined change in market rates and pricing levels. The
second table translates these yields or costs in dollars, based on expected
balances, which will reflect balances as forecasted by the business units
and adjusted for market movements and price changes.
Generating these matrices for each product and then aggregating them
would provide the retail bank with a dollar matrix representing the dollar
increase or decrease in income over the period considered. Dividing the
dollar figure by the average net asset size would provide the retail bank
with a Spread-at-Risk distribution. In other words, the Spread-at-Risk represents
the increase or decrease in the bank's net margin as a function of market
scenarios and the bank's pricing policies.
There are clear overlaps between strategic risk calculations for leveraged and non-leveraged financial institutions. Both require a sophisticated engine
to generate a large number of market scenarios efficiently, using one of
several methodologies that have gained acceptance in the market. Both require
the ability to model the "value" of a product or a transaction
as a function of market variables.
Yet, in the case of a non-leveraged financial institution, "value" designates the increase or decrease in the product's revenue or cost over
the period considered, as opposed to change in mark-to-market in the traditional
capital markets world. The concept of Spread-at-Risk thus provides non-leveraged
financial institutions with a more meaningful measure of its market risk,
as it integrates how both market movements and pricing policies can impact
an institution's net income over the horizon period.