•AR-Sen: A new CNN/ORC poll finds Republican Rep. Tom Cotton edging Democratic Sen. Mark Pryor 49-47 among likely voters, but in classic CNN fashion, the numbers among registered voters are wildly different. In fact, Pryor has a huge 47-38 lead with registered voters, suggesting that there's either a huge enthusiasm gap or that CNN's likely voter screen is too tight.
It's hard to say which is the case, though. In 2012, CNN's final poll of Florida found Obama leading by 6 with RVs while Romney was up 1 with LVs; Obama wound up winning by less than 1 percent, suggesting the LV model was closer to the mark. But in Colorado, the RVs had Obama ahead 8 while the LVs put him up just 2. Obama carried Colorado by more than 5 points, so there, the two voter screens split the difference.
If we have more of a Florida situation here, then Cotton either has a slim lead or the race is tied. But if things look more like Colorado, that would mean Pryor's the one in front. Steve Singiser compared RV and LV results for the final set of national polls in 2012: In every case, the RVs were closer to the mark than the LVs. And in 2004 and 2008, RV polls were also generally more accurate than LV polls. If this pattern continues, that's good news for Team Blue.
Pryor responded with his own poll (of likely voters, natch) showing him ahead of Cotton 45-43. The only prior survey from Hickman Analytics (conducted for a different client) came all the way back in February, when they had Pryor ahead 40-37. Note that Hickman's latest poll was conducted over a very long eight-day period that included Labor Day weekend, which is probably not the best time to get a read on the electorate. This is definitely not the only recent survey, though, that has this problem.