Brian, thanks again for your input on the calculator. To the best of my knowledge, you're the only person who has really dug into the code -- I really appreciate you taking the time to do that.
Firstly, I should note that this is my busy season with regard to my "real" work (i.e., the books), so it's likely that it may be a few months before I can really put much time into any major changes. (And, depending on what decisions get made here, it could be a major change.)
Regarding the "amount #4" not being added, thank you for catching that. That's just a plain-old error on my part. I hope to make that fix in the near future. (The fix should
be easy. It's the numerous affected tests that will take more time.)
As far as child mortality, you're right that ignoring possibility of child deaths does have the effect of significantly overstating NPV in adult disabled child scenarios.
It is generally a very minor effect on the actual recommendation though, because NPV is overstated very similarly from one selected filing age to another. That is, once the child begins receiving benefits, the amount that the child receives (either as a child benefit or as a surviving-child benefit) is generally unaffected by the age at which the parent filed for benefits. So after parent-age-70, the child benefit is the same regardless of when the parent filed. In other words, the NPV of two different filing age options will only be differently
overstated as a result of child mortality that occurs before the parent is age 70.
So if we only care about the recommendation (i.e., how good one filing age -- or set of filing ages -- is relative to another), then we're only concerned with how likely the child is to die between parent-age-62 and parent-age-70.
If we're dealing with minor children (rather than adult disabled children), probability of mortality in any 8-year window prior to age 18 is very low -- well below 1%
. For adult disabled children, it obviously would depend on how old they are, as well as probably depend significantly on what the disability is.
In writing the code, I was comfortable with assuming that a) people primarily care about how one filing strategy compares to another rather than the actual NPV and b) the child's death probability over the 8-year window is low enough that we don't have to worry about it throwing off the strategy comparison too much.
But either of those assumptions is certainly open to discussion. I'd be interested to hear what anybody else thinks on the topic.
(FYI I have jury duty tomorrow and am therefore unlikely to reply until Tuesday at the earliest.)