Thread by @Nate_Cohn

We have some good news for Democrats in North Carolina, where a reporting error by Randolph County is padding Trump by about 30k in our estimate there.
They reported both in person (abs one-stop) and absentee mail together, but the needle's still expecting 45k early votes

That's worth a net-30k votes for Biden, which would not be enough to get him into the lead. But it does help him a bit. We're looking for other errors

We also have a similar case in Rockingham County, which is worth another net-10k to Biden--still leaving him quite a bit short.
In the other direction, Orange County (heavily Dem) may have a bug on our end in expecting more IPEV there--but that's not so clear to me

If we add a net-40K to Biden, Trump would still be a heavy favorite but the needle would no longer give Trump >95% chance, because it does still believe there are late mail ballots left to count over the next few days--and we have no idea how many

The Orange County thing is legitimately odd. The county reported fewer early votes (heavily Dem here) than the needle expected, based on the state's absentee file (which was right elsewhere). So it's still thinking there's additional vote here... even though there shouldn't be

This could be a bug on our end somewhere, but I don't see why it would be. I think it is a legitimate possibility that there are another 8k votes here

This is actually something the needle is pretty useful for! Identifying possible reporting irregularities is... kind of important in close elections

https://twitter.com/tlaskawy/status/1323832093001031680?s=20

Fixing this Rockingham/Randolph problem is kind of tricky because... it might mess us up again if they fix it

Anyway, for now the needle *still* reflects the incorrect estimates for the early vote. I'll tell you if it's done, but if you're not sure you can just check our 'votes left' map for 'early' votes. I don't think either of the two big red bubbles or the blue bubble should be there