Residuals are a fact of life in attribution. Even for the most liquid securities, it is most unlikely that there will be a perfect match between the price as calculated from market curves, and the market price.
The consumer of attribution reports should therefore be deeply suspicious of a report that claims zero residual return. On many occasions, residual return may be redefined as credit return, but this should always be made clear to the user.
The best way to treat residuals is to accept that they convey useful information and that they act as a check that pricing accuracy lies within acceptable limits.
A very useful technique for business-as-usual reporting is to use an unattributed return bucket on attribution reports.
New securities enter portfolios all the time. Usually, these are based on existing patterns, such as corporate bonds or swaps. However, particularly in the fixed income markets and for larger investors, new security types may often need to be included in a portfolio even if they are not types that are included in the current security library, and so cannot be priced by the attribution application.
In this case the analyst has three courses of action:
Essentially, this approach defers solution of the problem. As long as its performance contribution is small, the presence of this new security will not substantially affect the overall conclusions of the attribution report, which tells the story of which global risks affected the portfolio’s absolute or active returns. Attribution reporting can then proceed as usual until the security can be modelled in the correct form.
In most cases, the third approach will be preferable to halting the production of attribution reports until the security has been classified.