[discount_rates_problem] [main_variables] [experience_bases]

There are 6 different experiences, for which random numbers have been generated. Of all of these possibilities, I believe that “twin_2019_indiv” (F) is a good option.

Whole_2018 [A]  This uses from end-1953 until end-2018 (formerly called “2018a”).

Whole_2019 [B] This uses from end-1953 until end-2019 (formerly called “2019a”).

Early_2019 [C] This uses from end-1953 until end-1984.

Later_2019 [D]  This uses from end-1985 until end-2019.

Twin_2019_Whole [E]  Using a simplified twin regime approach, “early_2019”  and “later_2019” were weighted 75% newer, 25% older, with the resultant random numbers then benchmarked to “whole 2019”.

Twin_2019_Indiv [F] This is the same as "twin_2019_whole" without the final benchmarking (effectively remaining benchmarked to sub-periods).