The range of bonds is checked every day and constantly updated to ensure the most accurate supply of data is being used. Any "stale" bonds are deleted and replacements sought. This raw data is error checked for consistency of rating, completeness, staleness of price, accuracy using standard deviations, comparison with historical data and bid/offer spreads. Any "rogue" data is fully inspected and adjusted or omitted as appropriate. "Out-takes" occurring during the process are retained in the database so as to allow auditing of the process.

The resulting spread(s) is reviewed against the underlying and where there is doubt concerning the spread it is either omitted or recalculated against a more suitable index.

The validated data is then used to populate a series of time charts and graphically represented to allow a further visual validation. This enables us to more closely inspect the interrelation  between issuers and maturities. The data is initially plotted into one month time buckets thus ensuring that, for example, a spread for a bond with 3 years and 3 months to maturity is differentiated from a bond with a term to maturity of 3 years 9 months.

Matrix by matrix and rating by rating, the details are plotted into more detailed versions of the matrices which are then populated with data so that each maturity node and rating receives a suitable spread. We use various interpolation and extrapolation methods to complete the matrices even for ratings where no freely available information exists. In some sectors, for example, there are few bonds available in say the longer maturities but by using correlations between other more complete matrices we are able to derive spread information for these missing ratings.

1.                  Only standard debt instruments are used for the initial database, e.g. bullet repayment, plain vanilla fixed and floating. Non-standard instruments, such as small issues, subordinated, high coupon or callable bonds are excluded from the spread database although we do retain the data for information purposes and back-testing of Liquidity Factors.

2.                  Ratings are checked for consistency and averaged with weightings applied to reflect issuers on Credit Watch. Where the three rating agencies we use disagree on the effective rating we recalculate the ratings to produce a weighted average rating. Where no issue rating is available we may sometimes apply an issuer rating.

3.                  Prices which have not been updated for more than 2 days are excluded from the database even though their implied spreads might still be relevant

4.                  Temporary aberrations in pricing are also removed from the database for that day.

5.                  Where implied spreads deviate substantially from the mean these are examined individually and a decision made as to whether they should be retained in the database or temporarily removed.

6.                  Where instruments have been removed from the main database they are not deleted but retained separately. They continue to be checked on a database and decisions made as to whether they should be re-included on a case-by-case basis.The underlying database is regularly reviewed and updated to capture new issues.