The covered bond directive has a surprisingly large number of paragraphs that make normal people question the intent. In the first of what I fear might be a long series of posts I have a bit of a rant about the treatment of interest in over-collateralisation calculations.
The two most annoying clichés of the covered bond market have to be ‘if it ain’t broke don’t fix it’ and ‘the devil is in the detail’. These clichés come together nicely when you spend a bit of time reading the ‘final, final’ text of the directive (as opposed to the ‘final, only minor technical changes from now on’ version).
The very first principle of covered bonds is that the liabilities should be covered by the assets. We then complicate things by stipulating rules about over-collateralisation and calculation methodologies – and there is plenty to debate in the new directive on those points – but the basic coverage principle has to be straightforward, surely?
Sadly not. Article 15 of the new directive states the general principle then defines the liabilities and the assets that have to cover them. The definition of liabilities has two problematic components. Firstly, interest on outstanding covered bonds.
“15(3)(b) the obligations for the payment of any interest on outstanding covered bonds”
That isn’t the amount of accruals of the next coupon (which would make sense from an accounting perspective as long as accruals on the underlying assets are also included), or even the whole amount of the next coupon (which makes sense from a prudential perspective and is netted against expected cash on assets). No. the wording implies all of the interest due on the outstanding bonds forever. So a ten year bond will have to cover twice as many coupon payments as a five year bond.
The definition of assets does not have a corresponding ‘payments of interest’ term in the formula but the article does go on to say that it is up to member states to define what this actually means. This could, at a stretch, be interpreted as allowing a member state to assume that the asset includes all future coupons even if the asset and liability wordings are asymmetrical. But that has two huge problems: with very few exceptions the assets and liabilities have different maturities: a 30 year mortgage funded by a 5 year bond has 25 years of coupons that it can take into account when calculating the coverage. And most mortgages have rate resets built into them either on a 3-month floating basis or at the end of a 2 or 3 year deal. So it is impossible to predict future payments.
Secondly, both the asset and the liability sides include their respective half of any swaps. For an interest rate swap this is obviously meaningless as the two notional amounts net off. Including both notionals implies that 200 of liabilities must be covered by 200 of assets. Silly but harmless, unless you also want to use this formula to calculate an over-collateralisation ratio in which case 105 of assets for 100 of liabilities becomes 210 of assets for 200 of liabilities – de facto double the amount that it would have been if this article had been better worded.
For foreign exchange swaps the problem is effectively the same to the extent that the can only be used to hedge exposures that are in the pool anyway.
The national discretion allowed in the calculation method might be sufficient to overcome the flaws in the wording but it would have been far easier to get it right in the first place, surely?