Trying to assess the required discount rate adjustment, I have always taken account of numeric values. In other words, in comparing the initial assumption (X) with the required initial assumption (Y), I have been looking at the difference Y-X. However, thinking about it, I think that I should have been looking at absolute values. The rationale is that it's as bad to be too far under as too far over.

Suppose that we have 100 adjustment “random readings”, of which 50 are -2 and 50 are +2. The mean is zero, apparently a great result. However, every single value was nowhere near zero so that the mean is not at all persuasive.

As an example for required adjustments for equities (“blend”, inflation-linked), the mark-to-market numeric mean  was 3.40% and the off-market numeric mean was -0.48%, suggesting that off-market is superior to mark-to-market. If, however, we used absolute values, those values would have been 5.19% and 5.71%, suggesting the opposite, albeit marginally.

Using the numeric values understates the errors, with absolute values being more informative. However, although I’m showing the absolute values in the “absolute adjustment” table, I’m sticking with numeric results for now. Please feel free to challenge me.