With the DNC starting tomorrow, this seemed like a good time to check in on the state of the Presidential race, looking at the most salient aspects based on available data.
Princeton Election Consortium: 93 percent probability Biden win
Economist: 89 percent Biden win
538: 72 percent Biden win
Electoral Vote Forecasts:
Princeton Election Consortium: 356-182 Biden
Economist: 350-188 Biden
538: 325-213 Biden
There are many other EV assessments outside of the modeled forecasts. They generally have Biden over 270 because they see him taking back MI, PA and WI. Assignment of other states varies and the overall split is affected by how many states they don't call but put in a "toss-up" category.
National Biden-Trump Trial Heat Averages
CNN: 52-41 Biden (+11)
538: 51.2-42.5 Biden (+8.6)
RCP: 49.8-41.9 (+7.9)
State Biden-Trump Trial Heat Averages
Too much data to list here. But places where you can get trial heat averages for states of interest are 538, RCP, CNN and CBS Battleground. Currently, in 538 Biden is ahead in every state Clinton carried in 2016 plus: AZ, FL, MI, NC, OH, PA and WI.
National Vote/Electoral Vote Split
This remains a real possibility but probably less of one than in 2016. Evan Scrimshaw remarks at Decision Desk HQ: "[Nate Silver's] estimates, at the launch of the model, say that Wisconsin is the tipping point state, and that a 1.9% Democratic lead nationally would be enough to win the state, down from the 2.8% it would have taken in 2016. In other words, Trump’s vaunted Electoral College advantage is now small enough that a Hillary Clinton margin of popular vote victory would be enough to win the Electoral College this time – at least, more often than not. A 2% Democratic lead is a ways away from the current high single digit averages, and even if there has been some recent movement, it is a lot easier to close a lead from ~11 to ~9, and incrementally harder to move it every unit after." Harry Enten notes: "If the live interviews polls [are] exactly correct, the difference between the tipping state in the electoral college (i.e. state containing the median electoral vote plus one) and the national vote would be under two points and perhaps closer to one point. It was nearly three points in 2016. A slightly under two-point gap is much more in-line with what has happened historically."
There was significant polling error in key states in 2016, primarily because state polls that did not weight by education undersampled white noncollege voters. That has generally been corrected this year. Also,there is little evidence that large numbers of "shy Trump voters" are being missed by polls, as a number of recent pieces have noted. Finally, with the kind of lead Biden has, any plausible polling error would not affect the outcome. Of course, if the race gets very, very tight, anything is possible.
Various indicators suggest that turnout should be very high this year. Michael McDonald, a leading expert on voter turnout and director of the US Elections Project says, ". “I expect voter turnout to be exceptional, perhaps the highest in over a century, since 1908." It's very difficult to gauge how much Trump's obvious efforts to interfere with mail balloting, which is expected to skew Democratic, might affect turnout, as opposed to simply delaying the vote count. Mail balloting can also have higher spoiled ballot rates, which might also have some effect on ultimate Democratic totals.
Recent polling suggests that black, Hispanic and Asian support for Biden is firming up, where previously it had been running a little behind Clinton's 2016 margins. Turnout of these groups is difficult to gauge in advance, but it would not have to be particularly robust to surpass that of the 2016 election. Also, it is worth noting that even if nonwhite turnout and support did not budge over 2016 levels,, Biden could still carry MI, PA and WI simply on the basis of underlying shifts in the population of eligible voters. Thus what is essential for Biden is less exceptional performance among nonwhite voters than reasonable performance among these voters, coupled with a Trump failure to expand his margin among white noncollege voters.
White Noncollege Voters
Biden is running 10-15 margin points ahead of Clinton among this demographic. Since Trump probably loses the election if he fails to expand his 2016 lead (+31) among these voters, Biden's success in significantly cutting Trump's margin, if maintained, is a death knell for Trump's chances. As an illustration, since nonwhite college voters will be such a large group in the 2020 electorate, if Biden cuts Trump's advantage by 10 points among this group (5 points more for Biden, 5 points less for Trump), that shifts the election 4.2 points in Biden's direction nationally. The effect would be larger in midwest states like MI, PA and WI where the white noncollege population is significantly larger. In contrast, increasing black turnout by 5 points (assuming no other group's turnout increases) would add .8 percentage points to Biden's margin nationally, more or less in specific states depending on how large their black eligible voter share is.
White College Voters
Polls indicate Biden is running ahead of Clinton's margin (+7) among white college voters by perhaps 5-10 margin points. Pew had Biden's margin at +23 among this demographic but I suspect this is an outlier. As a rule, any given improvement in Biden's margin among white college voters is worth about 30-45 percent less (depending on state) than an equivalent improvement in his margin among white noncollege voters.