[discount_rates_problem] [base_data]

For UK alone, I have looked at 8 different financial statistics series, covering all-share equities, long conventional gilts and long index-linked gilts. Each had a return over 1 year and a yield, making 6 so far. There are also 2 inflation series, over 1 year and over 15 years, making 8 in all. As RPI has a longer true data series than for CPI or CPIH, I have stuck with RPI.

The financials have been modelled annually from end-1953 until end-2018 and end-2019 (“2014” was used for my original 2017 DWP submission). For end-2019, I have then split the data between end-1953 until end-1984 (“early”) and end-1985 until end-2019 (“later”), taking account of when index-linked gilts became a more mature market. The base data used are summarised for means and standard deviations. The generated random numbers are explained here.

Because there were no data for the earlier period, the long index-linked gilts have only been modelled for the later period. Any comparisons involving index-linked gilts are really meaningful for 1985-2019 alone.

We have 4 experiences, namely “2018a” (1953-2018), “2019a” (1953-2019), “early” (1953-1984) and “later” (1985-2019). For each period, the random values have been adjusted to have the same mean and standard deviation as the base data.

A simplified “twin regime” approach was then used to build two further experiences, taking the “early” values 25% of the time and the “later” values 75% of the time. This is NOT the same thing as just weighting the values 25:75. So “2019b” takes those values and adjusts the mean and standard deviation to the base data for 1953 through 2019, while “2019c” omits that adjustment exercise. Using different names, they are mapped as follows:

1953-1984       Early_2019
1985-2019       Later_2019
1953-2018_a   Whole_2018
1953-2019_a   Whole_2019
1953-2019_b   Twin_2019_Whole
1953-2019_c    Twin_2019_Indiv