From the midweek edition of the Morning Jolt:

Yes, Trump Could Win. Emphasis on ‘Could.’
Words I don’t type often: salute Jonathan Chait. He’s seeing the race pretty clearly – or at the very least, he’s not telling Democrats what they want to hear.
The website FiveThirtyEight gives Trump more than a four-in-ten chance of actually, for real, winning. The Upshot, the New York Times’ forecaster, puts it at a slightly more comforting one in four, which sounds low except that, as the model’s authors point out, this makes the odds of a Clinton victory about equal to an NFL placekicker’s chances of making a 49-yard field goal. Also, the kicker has pneumonia. Citigroup recently warned that investors are underappreciating the significant risk to the economy of a Trump victory. A Trump presidency has felt unimaginable all summer long for the same reason Brexit couldn’t pass in England and Trump couldn’t win his party’s nomination: We refused to believe what the numbers were telling us…
The battle lines do not resemble those of a race between a regular politician and a freak-show buffoon. They reflect the normal Republican coalition, minus heavy defections from college-educated white women, against the normal, slightly larger Democratic coalition, minus the usual enthusiastic youth support. At the official level, public GOP defections are confined mostly to a handful of blue-state politicians and prestige print pundits. The party’s political and messaging apparatus, from Fox News to talk radio, remains pro-Trump, albeit often with muffled reservations.
With all of that in mind, it’s worth noting that Trump was on a really hot streak in polling for a while, but that streak appears to have cooled a bit. Nationally, most of the surveys show a small lead for Hillary Clinton or a tie. Clinton might be retaking the lead in Florida. While Trump’s numbers jumped almost everywhere, he’s still consistently trailing in Pennsylvania. His lead in Georgia is consistent but pretty small.
That’s the bad news for Trump. At this moment, Ohio looks good for Trump, Iowa is looking increasingly solid, North Carolina’s looking better and while Clinton still leads in Virginia, that lead is looking smaller. The one poll in Colorado this month had Trump ahead by 3, and if that isn’t an outlier or odd sample, then it’s a whole new race, because Clinton had been crushing Trump there earlier in the year.
Echoing the assessment of our Tim Alberta, Trump just doesn’t have any margin for error. He needs North Carolina and Colorado and Florida and Ohio and Iowa and Nevada.
Remember Sharron Angle led the final polls in Nevada in 2010 over Harry Reid; the unions had their get-out-the-vote efforts in the highest gear and Reid won by almost six points. Trump needs his turnout machine – largely outsourced to the Republican National Committee – to be hitting on all cylinders in every key state on Election Day to have a chance at 270 electoral votes.