the right answer is that it's tricky, but i need to point to a factor that the poll aggregates are not designed to take account of, and that's the undecided vote. i frequently root my criticism in polling in how it treats undecideds, so this should sound familiar if you've been listening in over a long period.
so, if you sort through the polling in ohio, for example, you'll see that the numbers are pretty consistent for biden - he's very close to 45%, across the board, in poll after poll. by contrast, trump's support modulates wildly from the low to high 40s. and, the data is not that different in these other places, either.
i'm used to analyzing this in canada, where the conservatives numbers just simply don't change, and polling is often distorted by ignoring undecideds. a frequent thing that happens in canada is that liberal supporters move to the undecided column in large percentages, which inflates support for the conservative party. so, you end up thinking conservative support is up, when the truth is that the sample just got distorted, because they removed the undecideds from the tally. when you put the undecideds back in, you always reconstruct that stagnant conservative support level. that means that the reality is that, in canada, the liberals are often their own most powerful opponent - they lose more support to themselves than they do to any other party,
in the united states, the republicans are the default, and you're sort of seeing something similar with trump. just because republicans don't like trump doesn't mean they'll actually vote for biden - they're still republicans, remember. they might stay home, or they might relent at the last minute. but, what i'm seeing suggests is that there's a fairly large group of people - more than 5% - that are struggling with their decision, right now. they're having a hard time telling pollsters they're going to vote for trump, but they don't seem to want to vote for biden, either. this is consistent with the broader data that suggests that there's a substantial amount of the republican base that doesn't seem to like donald trump because he's too liberal, something we've more often seen on the other side of the spectrum recently (democrats that think the nominee is too conservative) and that works itself out in canada with perennially disgruntled liberals that are pissed off that the party won't do what they say but that don't like the other parties much, either. they're coming in right now as undecided, but what that means in the american spectrum is "disenfranchised republican that doesn't know what to do" - just as it means "irritated liberal that doesn't feel they have a good choice" in the canadian spectrum.
so, it's hard to say.
is that what happened in 2016? in a word: no. what i'll say is that we saw something like this in some of these places very early in 2016, but it corrected itself quickly. that is, finicky republicans that didn't like trump only needed a small dose of hillary clinton before they came bolting back home - like a yappy dog given the option to run away. the party won that game of chicken, four years ago. the data is more persistent, this cycle, but the fact that they're still not picking biden is suggestive that they're perhaps not likely to, in the end.
what it means when you're reading the polls in the red states (the sunbelt states), you need to ask the question: is biden actually up or is trump just way down? if the latter, those polls may be misleading, in the end, as the undecided republicans start to come back. but, they're running out of time; conversely, if biden starts ticking up in these states, then trump's chances to win back wandering conservative republicans begin to find.
the caveat is florida, and that has to do with the fact that trump is increasingly losing seniors, which is making florida more like the great lakes states than the sun belt states, this cycle. in the great lakes states, the swing vote is traditionally democratic, and far more liberal than in most of the country. its also very old and very white. again - i don't think it's rational, but older white voters in the great lakes are leaning more towards biden than trump, and that is why biden is winning these states. the same thing is happening in florida. so, you may, oddly, see florida vote with wisconsin this cycle - and that may be decisive.
but, i need to continue to pencil in trump as having an advantage in these sun belt states until i see biden's numbers explicitly go up, rather than watch trump's numbers slowly tick down - and i need to argue that aggregate polling that gives biden an advantage as a consequence of these types of polling results is misleading, and likely in error.