One of the most arresting polls of the campaign, all the more so because it jibes with the data from Miami-Dade that Ed wrote about earlier. This isn’t some freakish outlier. Latino voters really do seem to be in play in Florida.

A paradox of this election is that it’s typically viewed as a battle of the bases. Can Democrats turn out more Trump-hating lefties than Republicans turn out Trump-loving righties? But the hidden story of the race is that each candidate is poaching key groups from the other party. Typically senior citizens are a Republican group. In Florida four years ago, Trump crushed Hillary 57/40 among them — the difference in his narrow win statewide. This year Biden has been surprisingly competitive among seniors, even leading in that demographic in many polls. Including here: NBC/Marist has him up 49/48 in Florida among the 65+ group. If Trump were duplicating his 2016 margins with seniors, he’d be leading in Florida.

So how is it that Marist has the state tied overall at 48 percent apiece? Simple: Trump is stealing votes from Biden’s base too. Four years ago he lost Latinos in Florida to Clinton by 27 points. Today? He’s up four. A 31-point turnaround!

Biden’s biggest advantages in the tied Florida race are among Black likely voters (he leads them 83 percent to 11 percent), women (57 percent to 40 percent) and independents (51 percent to 40 percent).

Trump, meanwhile, leads among men (58 percent to 38 percent), white voters (56 percent to 41 percent) and whites without college degrees (63 percent to 35 percent).

Yet in a significant break from four years ago, Trump holds the narrow edge over Biden among likely Latino voters, 50 percent to 46 percent — with Trump leading sizably among Latinos of Cuban descent, and with Biden just slightly ahead among all other Latinos in the state. (The Cuban-versus-non-Cuban Latino split in the poll is too small to report the actual numbers.)

The gender split there is almost as interesting as the split among Latinos. Typically Trump leads comfortably among men while Biden leads enormously among women, an imbalance that favors the Democrat. That’s not what we see here, though; Trump’s margins among male voters exceed Biden’s among females. And meanwhile Latinos aren’t the only minority that are a bit more open to Trump than they were four years ago:

Be careful of the small sample sizes, but this is a pretty consistent picture we're seeing in plenty of polling… This is from FL comparing Marist's final poll in 2016 and their new 2020 poll. pic.twitter.com/tmIMPNjlpz

— (((Harry Enten))) (@ForecasterEnten) September 8, 2020

Trump’s cutting into Democrats’ margins among blacks and younger voters while Biden’s cutting into Republicans’ margins with whites and seniors. Whoever carves off more of the other guy’s base is likely the next president.

As for the Latino vote, theories are flying:

Pure conjecture: Seems possible that Trump’s anti-socialist messaging is particularly persuasive among older Cubans in Florida.

— Derek Thompson (@DKThomp) September 8, 2020

Right, but Trump’s been tossing bombs at Biden and the Democrats for being socialists since the primaries. Didn’t do him much good until now, with Sleepy Joe enjoying big leads in the state over the summer. On July 25, FiveThirtyEight had him ahead by an average of 7.6 points in its poll tracker. Today?

An obvious possibility: As the pandemic has slowed down in Florida, voters there have been less consumed with assessing Trump’s performance on containing the virus and more focused on news coverage of left-wing violence in cities. If that’s right then the presidency may boil down to how mild or severe the “second wave” of infections this fall is in the state.

This is worth keeping in mind too:

Here's the problem with overly focusing on Hispanics instead of the white working-class voters… While it improves Trump's chances of keeping FL, TX, AZ, it makes it harder in PA, MI, WI, OH, NC, and NH https://t.co/WYlYMassmX

— Ryan James Girdusky (@RyanGirdusky) September 8, 2020

Biden’s outperforming Clinton among whites and whites remain the key in the “blue wall” states that turned red four years ago. If Joe were to hold all of Hillary’s states (which is a big if given how tight Minnesota is right now) and flipped those three, he’s the next president no matter what happens in Florida. On the other hand, the polling in Pennsylvania doesn’t look so hot lately for Biden either:

In the current survey, available at www.susquehannapolling.com, and attached to this release, Biden’s lead has been cut from 6 points in April (48:42 margin), to 5 points in June (46:41), to now a 2-point margin. Momentum for Donald Trump in the current poll seems to be the result of a surge in voters who believe “restoring law and order” to U.S. cities facing looting, rioting and violent protests should be a top priority for federal lawmakers. For instance, when voters are asked to choose the top two issues (from a list of 7) they want federal lawmakers to concentrate on, “restoring law and order to cities and communities” is chosen by 37%, ranking this issue in third place among all issues and nearly tied with “improving access and affordability to health care” (at 38%) and “strengthening the economy and job creation”, ranking highest at 40%. Perhaps most significantly, not only does restoring law and order to U.S. cities outrank concern among those who believe improving race relations in the country should be a top priority (tied for 4th place at 21%), but “law and order” voters prefer Donald Trump over Joe Biden by a 74% to 16% margin. This is the clearest evidence yet that violent protests and rioting across the USA are hurting Biden’s chances in a must-win state like Pennsylvania. Conversely, voters who believe “improving race relations” should be a top priority are voting for Biden over Trump by a 78:11 margin.

Is that a fluke result? Not really. Quinnipiac found Biden comfortably ahead in Pennsylvania at the end of August but all other polls taken there recently have been tighter. Monmouth and Change Research each had the Democrat leading by three and Rasmussen had the race tied. Both the RCP and FiveThirtyEight poll averages of the state have Trump down just four points.

An interesting footnote in the news today: Biden is beefing up his staff in Georgia, not a state you’d expect to see him focused on as the polls tighten in Florida and Pennsylvania. It’s hard to believe he’d have a chance there if other swing states are starting to tilt against him. (The latest poll taken of Georgia had Trump up seven.) My guess is that he’s leveraging his gigantic fundraising windfall in August to try to force Trump to play defense there at a moment when Republicans’ money advantage has disappeared. Every dollar he can make Republicans spend across the border in Georgia is a dollar they’re not spending in Florida. It can’t be that Biden is trying to expand the map, though, at a moment when his advantage in more winnable swing states is shrinking. Not unless his polling of those states looks a lot different from the public polling.