The paper linked to by aerophilic goes into that issue in some detail, but it does not lead to a sharp distinction, either - for one thing, the ballistic coefficient ranges over a couple of orders of magnitude, with smaller objects coming down sooner.
If we confine the discussion to circular orbits, it suggests that 100km is a better number than anything lower, but elliptic orbits can get away with lower perigees, which is a concern when the boundary of space is being defined for national territory purposes (e.g. spy satellites with a perigee over the nation being spied on.)
If we confine the discussion to circular orbits, it suggests that 100km is a better number than anything lower, but elliptic orbits can get away with lower perigees, which is a concern when the boundary of space is being defined for national territory purposes (e.g. spy satellites with a perigee over the nation being spied on.)