Question ID: 660
Regulation Reference: Risk-Free Interest Rate - Matching adjustment
Article: 43
Status: Final
Date of submission: 29 Mar 2016
Question
It appears that the Matlab code publised contains a bug in the calculation of the LTAS for sovereigns other than EUR countries.
The technical specs and the Q&A in the reference seem to suggest that "The LTAS for sovereigns is derived by comparing the historical yield on sovereign bonds with a retrospective construction of the Solvency II basic risk-free interest rate. "
Instead, the Matlab code loads the Risk Free Rate excluding CRA for the currencies other than EUR, making the LTAS always equal the CRA for the specific currency.
Could you please advise which is the intended algorithm? If the Matlab code indeed contains an error, is this going to be corrected and the LTAS calculated based on the correct yields and in turn spreads?
EIOPA answer
We have checked the calculation of the LTAS for government bonds in the Matlab source code. Our conclusion is that for all currencies the LTAS is based on a comparison of government bond yields and basic risk-free interest rates. The basic risk-free interest rates are adjusted for credit risk. This is the intended calculation.
The LTAS calculations for government bonds are in the code of the file “RFR_08A_LTAS_Govts.m”. The history of basic risk-free interest rates is loaded in line 248 of the code. The basic risk-free interest rates includes the CRA. The history of government bond rates is loaded in line 316. For the relevant dates the differences (spreads) are being calculated. These outcomes are used to update the LTAS according to section 10.B.1 270 of the technical documentation.
Please note that where the basic risk-free interest rates are derived from government bonds, the difference between government bond yield and basic risk-free interest rate equals the CRA. In this case the LTAS is an average over CRAs.