That road leads to "how many golfballs in a minibus" type of puzzlers and we already have data now that shows that those things don't work.
http://www.deathandtaxesmag.com/200732/google-admits-its-fam...
http://www.deathandtaxesmag.com/200732/google-admits-its-fam...
I mean actual as in 'problem that you actually have and that needs solving'.
OTOH if you feel that the problems you are dealing with are solvable (and they had better be) then working together with your new prospective hire on some aspect of them will quickly give you insight into how they would go about this and whether or not they have something to contribute.
It's by no means a super necessary skill to have though. And certainly it makes for a rather silly interview question. I would guess it takes no more than half an hour to an hour to train yourself to do estimation problems.