The new measures are calculated in the same way as the old ones, but instead of weighting together annual percentage changes, seasonally adjusted monthly changes are weighted together, annualised. The seasonal adjustment is based on data from 1980 onwards. An implication of this is that the seasonal adjustment is re-estimated for the entire history each time new data are added and that the history of the measures changes for each new month of data added.

TRIM1 and TRIM85 are calculated by assigning the subgroups with the highest and lowest seasonally adjusted monthly percentage change in a given month a weight of 0 for that month, while the other subgroups are weighted up so that the weight sums to 1. Thus, different subgroups will be excluded in different months. In TRIM85, subgroups are given a weight of 0 when they correspond to a weight sum of 7.5 per cent and have the highest annual percentage change. The same applies to the subgroups that correspond to a weight sum of 7.5 per cent and have the lowest annual percentage change. However, there will always be subgroups that are right on the edge. Instead, the weight for that group is adjusted depending on how much of the weight total needs to be excluded to reach a total of 7.5 per cent. Then all remaining weights are weighted up so that the weight sum is 1. These weights are then used to arithmetically weight the annual percentage changes for the different subgroups. For TRIM1, the procedure is the same but now all but the middle 1 per cent are removed.

The UND24 measure retains all subgroups but gives them a different weight than in the CPIF. The weights are calculated by first calculating the difference between the monthly percentage change in each subgroup and the change in the overall CPIF. The weight for each subgroup is then calculated each month based on the historical standard deviation of the deviation. More specifically, the weights are calculated by first calculating the inverse of the 24-month moving standard deviations of the deviation for the different groups. These are then normalised so that the sum of the weights is 1 in each time period. These weights are then used to arithmetically weight the annual percentage changes for the different subgroups. The weights thus vary from month to month and the weight is larger for the groups where the variation has been small compared with total CPIF inflation, while it is smaller for the groups where the variation has been large compared with total CPIF inflation over the past 24 months.

A similar approach is used for the CPIFPV measure. Again, all subgroups are retained. But here the weights are determined by the persistence of the monthly percentage changes for each subgroup. This is done by estimating a simple first-order autoregressive model on each subgroup. The weights are then obtained by normalising the coefficients for each subgroup so that the total weight sums to 1. These weights are then used to arithmetically weight the annual percentage changes for the different subgroups. The estimates are based on a rolling window corresponding to the last 60 months, i.e. a new estimate is made every month, which means that the weights change continuously. The weight for an individual subgroup will thus be greater the higher the estimated autoregressive coefficient for the monthly percentage change rates has been for the subgroup over the last 60 months.

The CPIFPC has been produced using what is known as principal component analysis. Again, all subgroups are included in the calculations. First, the subgroups are standardised so that all have a mean of zero and a standard deviation of 1 for the given period. Static factors for the subgroups are then estimated using principal component analysis. This is a method of trying to reduce the data set to a few components that can explain much of the overall variation in the data. We weight the first three components, which together explain about 15 per cent of the variation in all subgroups. The weight they are given in the weighting is based on how much of the total variation in all subgroups each component can explain of the total variation. Finally, a simple regression on the CPIF is estimated with the factor as the only explanatory variable; the CPIFPC is the fitted values from that regression.