Taking the normal approach as it is, we can calculate the “estimation error” as the initial estimate minus the actual outcome, looking at differences over time.

That is what we have done first, using the “over 15” gilt indices for the estimates against RPI over 15 years annually. Although, secondly, the errors have recently been somewhat lower than in the past, we would not want to rely upon that continuing into the future. It will be seen that the differences have sometimes been enormous (for example, 4.1% pa for 1990 to 2005), which is why we originally started seeking alternatives.Separately, on a monthly basis since the end of 1998 until 2022, we have shown how the error is built-up.

Comparisons between gilts, forward and spot can be found here.