I'm hoping you're just kidding around, because...margin of error.
Of course, I'm kidding, but just to extend this...
What is the margin of error on the "already voted" component? How does the polling outfit screen out the crowd who says they've voted but really haven't?
I only point this out because the previous poll that arrived at the 19% conclusion was a few days before the Ohio Secretary of State was reporting about 5.5% already voted (absentee + in-person early).
So unless the margin of error there is +/-13.5%...