We are comparing initial estimates with the actual outcome after 15 years. Referring to a year [X] means the period of 15 years starting at the end of December [X]. For gilts, we’re looking at the yields on stocks which are definitely redeemable after at least 15 years. The “spot” and “forw(ard)” estimates used come for duration 15 from the Bank of England yield curves. The initial estimate is simply the nominal measure minus the real measure. As a further alternative to “Bank of England”, we tried looking at deriving forward estimates from past experience for the same period length but the patterns over time were so chaotic that we saw no point in pursuing this line of thought any further. If you know of anything else which may be relevant, please do tell us about it.

Current gilts-based inflationary expectations just haven't been working well. They have mostly tended to be biased too high, with the converse in recent years. The standard approach is mandatory for PPF calculations (which are meant to be “prudent”) and merely recommended for FRS17 (or IAS19) calculations (which are intended to be “realistic”).

One point which should be taken into account much more often is that prudence can only be assessed against a best estimate. If the current general approach is hardly “best estimate”, it is hard to see how it can possibly be taken as prudent. We suspect it is actually “super-prudent”, directly leading to huge capital and social losses, in the form of wildly inflated deficit payments inflicted upon business stakeholders and losses of valuable pension rights by a huge number of employees.

The approach used has led to continuing significant corporate accounting distortions and we think this matters. Broadly, we believe deferred pensioner liabilities are being overstated by at least 25% and current pensioner liabilities by at least 15%, perhaps substantially even higher.

Just looking at the mean errors, which is insufficient, the gilts estimate for 1985-2010 was 0.99% too high, bettered by 0.67% for spot and 0.27% for forw. Over the first half (1985-1997), the estimates were in the same order but much higher, namely 2.18% (gilts), 1.72% (spot) and 0.71% (forw). Over the second half (1998-2010), all 3 did much better at (0.20)% for gilts, (0.39)% for spot and (0.17)% for forw.

Another approach is to look at how often the absolute value of the error was lower than 0.5% (arbitrary). For 1985-2010, with 26 values, the 3 measures were virtually identical at 10 for gilts, 11 for spot and 12 for forw. During 1985-1997, with 13 values, spot and gilts had just one such reading while forw, had 4. During 1998-2010, also with 13 values, the 3 measures were virtually identical at 9 for gilts, 10 for spot and 8 for forw. This is not hugely helpful.

Finally, we wondered which measure was closest to zero. For 1985-2010, forw was closest 16 times with 6 for gilts and 4 for spot. For 1985-1997, forw was closest 10 times with 2 for gilts and 1 for spot. For 1998-2010, forw was closest 6 times with 4 for gilts and 3 for spot.

All in all, without ignoring the first half of the experience, we still think that the “forward” measure is the best of the three for estimating future RPI movements over long periods. Over the whole period, the error was an over-estimate of a mere 0.27%, which we consider can be taken as zero.