On 12 May 2023, Henry kindly published a piece we wrote, querying why index-linked gilts had been so firmly recommended by TPR and consulting actuaries. No one has responded, which is interesting in itself. Anyway, that question was partly sparked by our joint website ukrpi.com, which we’ve been running for around 15 years, which we think worth describing.
For long-term DB pension funding, it has been very common for UK actuaries to estimate future long-term RPI inflation by comparing yields on conventional gilts (“CGs”) with those on index-linked gilts (“ILGs”). The point of our website is to show that this doesn’t work
Why have actuaries done that? Yes, we can see the “logic” but where, if anywhere, was that recommended or suggested? Having asked around, no one seems to know. Although we had thought it was mandatory for PPF §179 assessments, that is not the case (see original guidance from 2005). For FRS17, there is merely a strong steer in section 26, without any reference to duration for inflation. While IAS19 has mostly swept FRS17 away, we are unaware of any further relevant guidance. Given that yield curves tend to slope upwards, that does beg questions. While similar outstanding terms may seem logical, maturity may be more of a problem than has generally been observed. After all, the ILG “duration” must be longer than for conventional gilts because the return is very much more heavily weighted towards capital.
The rationale appears to be that gilts of either kind should deliver broadly the same return over a long period and that the difference represents the allowance for inflation. On performance, over periods of 15 years, ILGs have performed marginally better than CGs, by a mean margin of 0.3% pa. Over short periods, the out-performance has been the reverse. We thought we should test the inflation approach against reality, which we started in around 2008.
In fact, we have found that the standard approach hasn’t worked at all well. As an example, take the 15 years from Dec 1990 to Dec 2005. In 1990, the expected inflation was 6.8% pa, against actual inflation measured in 2005 of 2.7% pa, a colossal difference of 4.1% pa, the highest observed. Using annual readings between 1985 and 2022, the mean difference between estimate and actual was an over-estimate of 1.2% pa (standard deviation 1.4% pa).
We have also looked at what might be a sensible estimate of the difference between RPI increases and CPI increases over periods of 15 years. For at least several years, we have read and heard it claimed that the long-term difference is equivalent to 1% pa. Comparing RPI with CPI over 1 year, the average has been 0.78%, with a standard deviation of 1.07%. As it occurs at the 60th percentile, it is not unreasonable to suggest that the expected difference could be as high as 1%. However, comparing RPI with CPI over 15 years, the average has been 0.72% (very similar to that over 1 year), with a much lower standard deviation of 0.10%. A value of 1% would be 2.8 standard deviations away from the mean, which corresponds to a likelihood of 1 in 200, namely highly unlikely; 1% really cannot be taken as a best long-term estimate.
The website has a number of charts and we do look at CPI and CPIH as well as at RPI. Further, we have also tracked Bank of England yield curves, which, overall, seem to be rather better than using gilts in a simplistic basis. If anyone feels that we have misstated how actuaries estimate inflation, and how they have been generally encouraged to do so, please tell us.
LongTermInflationEstimates_ConAndJon_14Jul2023.docx
