Spread changes – historical data

Since the correlation between changes in credit spreads and changes in interest
rates is unstable (the correlation even changes sign over time), it is important to
measure a portfolio’s or an institution’s exposure to spread risk independent of an
assessment of interest rate risk. For example, a portfolio of Treasuries with an
effective duration of 5.0 has no spread risk, but does have interest rate risk. A
portfolio of corporate bonds with an effective duration of 3.0 has less interest rate
risk than the Treasury portfolio, but if adverse moves in interest rates and spreads
occur simultaneously, the corporate portfolio may be a greater source of risk than the
Treasury portfolio. Spread risk affects corporate bonds, mortgage-backed securities,
asset-backed securities, municipal bonds and so on, and a change in spreads in one
segment of the market may not carry over to other areas, as the fundamentals and
technicals that affect each of these markets are typically unrelated. Nonetheless, we
have seen that in times of extreme uncertainty, correlations across markets can
converge rapidly to ò1.0, eliminating the benefits that might otherwise be gained by
diversifying spread risk across different market sectors.
What magnitude of spread changes can one reasonably expect over a given period?
Table 7.1 shows the average and standard deviations of option-adjusted spreads for
various sectors over the six-year period, August 1992 to July 1998. In parentheses,
we show the standard deviations computed for a slightly different six-year period,
November 1992 to October 1998 (note: statistics were computed from weekly observations).
The reason the two standard deviations are so different is that the values in
parentheses include October 1998 data and therefore reflect the spread volatility
experienced during the market crisis discussed above. Of course, six years is a long
time and statistics can change significantly depending upon the observations used
to compute them, so we also show the average and standard deviation of optionadjusted
spreads measured over a one-year period, August 1997 to July 1998, with
the standard deviation computed over the one year period November 1997 to October
1998 shown in parentheses.
When evaluating the importance of spread risk, it is important to stress-test a
portfolio under scenarios that reflect possible market conditions. Although the
October 1998 experience may certainly be viewed as a rare event, if we use the oneyear
data set excluding October 1998 to forecast future spread changes (based on standard deviation), we could potentially underestimate a portfolio’s spread risk by
more than threefold (the average standard deviation based on one year of data including
October 1998 is 3.5 times the average standard deviation based on one year of
data excluding October 1998). When stress-testing a portfolio, it would be a good idea
to combine spread changes with interest rate shocks – for example, what would
happen if interest rates rise by X bps while corporate spreads widen by Y bps and
mortgage spreads widen by Z bps? If we have computed the portfolio’s effective duration
and the spread duration of the corporate and mortgage components of the portfolio,
we can easily estimate the impact of this scenario: [(Effective durationOverallîX)
ò(Spread durationCorporatesîY)ò(Spread durationMortgagesîZ )].