Finite population correction, election

classic Classic list List threaded Threaded
1 message Options
Reply | Threaded
Open this post in threaded view

Finite population correction, election

Rich Ulrich
On MSNBC, Steve Kornacki has been giving a pragmatic tutorial
on applying the Finite Population Correction to incoming reports
of votes, as the final counts of mail-in ballots are trickling in from
Pennsylvania, Arizona, and Nevada.  He was describing Georgia,
too, until the total count reached close enough to 100%, with a
tiny margin, that officials declared that there would be an automatic

Contrary to many applications of the FPC, the projection for final
counts is stratified, and uses knowledge about the "not-counted"
votes.  For instance, in Pennsylvania (my state), mail-in voting is
new, and half the voters decided to do it. But because Trump has
railed against mail-in voting for months, most of the Pa. mail-in votes
are for Biden - 76%, according to Kornacki, a few minutes ago. That
contrasts to the vote on election day.

Trump gained a lead of a half-million+ on election day when only the
same-day votes had been counted.  Mail-votes have shifted the balance
to a Biden lead of 27,000.  Some 100,000 of the remaining votes are
"provisional" -- a different category/stratum from "mail-in".  K says that
Trump needs to win 70% of those to close the gap (which is not likely).
We don't know how many mail-votes are yet to be counted, but K has
implied that the number could raise Biden's lead to 50 or 70 thousand
or more /before/ provisional ballots are included in the count.

Rich Ulrich
===================== To manage your subscription to SPSSX-L, send a message to [hidden email] (not to SPSSX-L), with no body text except the command. To leave the list, send the command SIGNOFF SPSSX-L For a list of commands to manage subscriptions, send the command INFO REFCARD