,
4 tweets,
1 min read
Read on Twitter
State polls in 2018: pretty good, but still off in ways highly reminiscent of 2016 nytimes.com/2018/11/21/ups… 
It's a little hard for me to look at this in our own polls, mainly because they're CD level. But tbh I don't really see much evidence of the same pattern in the upper-right (we did underestimate Dems in the southwest, tho)
To take one high-profile example: we're basically the only public poll that had Blackburn ahead (+14) by more than her eventual margin of victory (+12). One of our least accurate polls was a GOP-biased poll in MN08. Overall, no bias in polls of states where polls were D in '16
Basically, I think it would have been nice (strictly for this purpose) if we had polled Indiana at the end. We didn't in part because our reporting indicated Braun was ahead, which meant that the Senate wasn't interesting. Then Marist/Fox had late polls showing Donnelly up
