I think I agree with you Lou, but I'm trying to make sense of it. See if you
(and everyone) agree with this as I step through the procedure...
1) We start with premium for a current policy we are rating and we multiply by
the ELR to get an estimate of the losses of the policy being rated. These
expected losses are really an estimate of the FUTURE.
2) Since these are future estimates, we "untrend" these expected losses to
give an estimate of what they would have been during each of the 3 years in
the experience period.
3) Use LDF's on these past years expected experience (for each year
separately) to estimate the losses expected to emerge from this expected past
experience.
4) add the losses expected to emerge to the actual losses from the past.
If we had not untrended the expected losses, and had applied the LDF's to the
expected losses for the rated policy, we would have been adding development on
losses expected in the future (for the rated policy) to the actual losses we
collected on past policies.
5) Now we have the total "actual" losses from the past (actual losses
observed plus an estimate of losses expected to emerge on these past
policies), and we compare that to the losses we expected from these past
policies (the sum of the losses from #2 above). This gives us our AER.
Given all that, I would agree that the losses included in the AER are not
future losses, and thus I could say that they are not trended (the answer to
my 1979 question).
So, if there are no errors in my synopsis above, then I think I'm o.k. with
the answer given in the manual (that losses in the AER aren't trended).
Also, my two cents is that I don't think the question is moot, eventhough it's
old. I think this type of question could be asked.