There has been a lot of discussion lately about how the electorate has evolved from Donald Trump’s narrow victory in 2016 to his likely loss in 2020: Senior citizens have swung back in the Democratic direction, or Latinos are more likely to vote for Trump, for example. Some of that pundit chat has just been taken as an article of faith, though, originating with snippets from focus groups or person-on-the-street interviews, or extrapolated from looking at which states are moving in what direction and who lives in those states.
But there’s actually an illuminating and rigorous way to do that kind of number-crunching in an apples-to-apples fashion, by comparing the crosstabs from the same good pollster from four years ago versus today. And if you do that with Pew Research’s most recent national poll, it turns out that much of the conventional wisdom is in fact borne out: Trump’s Latino vote share has indeed increased, but that’s counteracted by continued Democratic improvement among college-educated white voters. And while Biden is doing better with senior citizens, his real gains seem to be coming with younger voters, a much less-discussed demographic.
In this post, we’ll compare the crosstabs of Pew's October 2016 poll with their October 2020 poll, including their separate component on religion. If you’re wondering why Pew, in particular, that’s partly because of a long track record of accuracy in horserace polling, but also because they tend to use much larger sample sizes than other pollsters, which means that the crosstabs themselves have large enough sample sizes to be considered reliable. Pew’s 2020 poll features, for instance, 11,929 respondents, who were validated against 2016 and 2018 voter files. By contrast, most routine opinion polls have sample sizes in the mid-hundreds. That might be adequate for accurate topline numbers, but it yields completely unreliable data when drilling down to individual crosstabs.
Speaking of topline numbers, that’s where we should start. The 2020 Pew poll has Joe Biden leading Donald Trump 52-42 nationwide, while Libertarian candidate Jo Jorgenson gets 4% of the vote and the Green Party’s Howie Hawkins takes just 1%. Their October 2016 poll had Hillary Clinton leading Trump 46-40, with 9 for Libertarian Gary Johnson and 3 for Jill Stein, the Green candidate.
Some folks might naturally ask, “Why would you use the 2016 poll? It was wrong!” Digging deeper, though, it really wasn’t that far off. Clinton, of course, did wind up winning the popular vote by a margin of 2 points, 48-46. Beyond that, it’s possible, if not likely, that the actual state of the race in mid-October of 2016 was something around Clinton 46, Trump 40. That was right before FBI Director James Comey’s infamous letter shifted the race’s dynamic, and it’s clear from panel-type surveys (see the graphic on page 7 of this link, for an example) that in the final weeks, the undecided voters who didn’t like either candidate broke heavily—in fact, almost entirely—to Trump, which explains how he pulled ahead at the very end.
You might say that polls in 2016 weren’t so much “wrong” but rather that most firms, especially state-level pollsters, just stopped polling too early to catch this late drift. And that also points to one of Biden’s key advantages this time, which doesn’t get discussed very much: This year, there are many fewer undecided voters—or voters weakly supporting third-party candidates, the kind who usually wind up voting for a major party candidate anyway at the end. You can see that in the many polls, both national and in swing states, that have Biden already over the 50% mark.
But let’s start looking at the various crosstabs, starting with race:
RACE
RACE |
CLINTON |
TRUMP |
2016 MARGIN |
BIDEN |
TRUMP |
2020 MARGIN |
CHANGE |
WHITE |
38 |
49 |
-11 |
44 |
51 |
-7 |
+4 |
(COLLEGE WHITE) |
51 |
36 |
+15 |
58 |
37 |
+21 |
+6 |
(NON-COLL. WHITE) |
31 |
56 |
-25 |
34 |
60 |
-26 |
-1 |
BLACK |
81 |
3 |
+78 |
89 |
8 |
+81 |
+3 |
HISPANIC |
65 |
18 |
+47 |
63 |
29 |
+34 |
-13 |
The most interesting data point may be that Trump’s position with Latino voters is, indeed, better than where it was four years ago. Note, however, that Biden’s standing with Latinos is not materially worse than Clinton’s four years ago; it may simply be that many fewer of them are undecided this year, compared with where they were four years ago.
A natural follow-up question to this would be: “Why? What about … all the racism?” Well, at least part of an explanation is that Cuban-Americans (and probably Venezuelans as well) seem to have warmed up considerably to Trump since 2016, very possibly based on his aggressive stances toward the governments in their ancestral countries. You can see that happening if you look specifically at polls of Miami-Dade County or Florida’s heavily Cuban 27th Congressional District, which provide some of the few surveys you can find anywhere where Trump is overperforming his 2016 stats. That’s probably only a partial explanation, though, because Cubans are only a small part of the nation’s overall Latino population and by themselves couldn’t bring about that big of a shift.
In the final equation, though, that Latino shift may not matter that much for Biden’s prospects, since it’s being more than canceled out by the movement in the opposite direction among white voters, especially college-educated ones. While that movement doesn’t appear as large (“only” 4 points among all white voters nationwide, 6 among college-educated white voters), it still has a bigger impact because white voters substantially outnumber Latino voters in most parts of the country. In other words, 4% percent of two-thirds of the total electorate (the plausible white share we can expect) has a lot more leverage than 13% of around 15% of all voters (the approximate Latino proportion).
Even states with large Latino populations like Arizona (which has apparently moved into Biden’s column) and Texas (which, regardless of the final outcome, has undeniably graduated into long-anticipated swing state status) are nevertheless seeing significant improvement in Democratic vote share. Again, much of that is because of college-educated white voters in those states’ suburbs fleeing the Republican Party.
That’s even what seems to be happening in Florida, where Biden is maintaining a lead in the low single digits despite what’s going on in the Cuban-American community, because he’s bouncing back so strongly in the suburbs of the Tampa and Orlando areas. As a case in point, mostly suburban Pinellas County, which Trump narrowly won in 2016, is now on track for a double-digit BIden win.
EDUCATION
EDUCATION |
CLINTON |
TRUMP |
2016 MARGIN |
BIDEN |
TRUMP |
2020 MARGIN |
CHANGE |
POSTGRAD |
64 |
25 |
+39 |
68 |
28 |
+40 |
+1 |
COLLEGE GRAD |
49 |
35 |
+14 |
57 |
37 |
+20 |
+6 |
SOME COLLEGE |
43 |
41 |
+2 |
48 |
46 |
+2 |
0 |
HS OR LESS |
40 |
47 |
-7 |
45 |
49 |
-4 |
+3 |
While the education gap is quite large, the growth in the education gap seems to have already happened, by and large, since the gulf isn’t getting much bigger this year according to Pew. The education gap really opened up in 2012 and especially 2016; it was was much smaller in the 00s, and in fact ran the other direction in the 20th century, when people with lower levels of education were more likely than people with more schooling to vote Democratic.
So it’s heartening for Democrats to see that Biden is bouncing back better across the educational spectrum, among non-college voters as well as college-educated ones. Looking at state-level and district-level polling, that bounce-back isn’t uniformly distributed, but among non-college voters, we’re seeing it more in areas lots of white voters but not many evangelical Christians. The states of the Upper Midwest are prime examples, but West Virginia (both one of the whitest and least-educated states) is maybe the most extreme case: Most polls of the state still show Biden losing badly, but by margins in the teens, versus Clinton’s 40-point drubbing in 2016.
AGE
AGE |
CLINTON |
TRUMP |
2016 MARGIN |
BIDEN |
TRUMP |
2020 MARGIN |
CHANGE |
18-29 |
49 |
28 |
+21 |
59 |
29 |
+30 |
+9 |
30-49 |
47 |
34 |
+13 |
57 |
37 |
+20 |
+7 |
50-64 |
43 |
47 |
-4 |
49 |
47 |
+2 |
+6 |
65+ |
45 |
47 |
-2 |
49 |
49 |
0 |
+2 |
There’s been a lot of talk about how Biden's rebound among senior citizens is what’s driving his success this year, but Pew’s data suggests that if any one age group is making a difference, it’s younger voters. Biden has still improved with the 65-plus set, but the younger you go, the bigger the gains. This is one area, though, where Pew seems to see things somewhat differently from other pollsters, some of whom have found a double-digit surge for Biden among senior citizens.
Interestingly, Trump isn’t doing any worse with young voters; the difference maker is that Biden is polling at 59% with them, instead of 49% as with Clinton. In other words, many more younger voters have already committed to Biden instead of calling themselves undecided or holding out for a third party.
We may have a less contentious Democratic primary, and a quicker concession from Bernie Sanders this time, to thank for that. (It’s possible that each new generation of young voters has to learn the hard way not to experiment with vote-wasting, just as the previous generation found out with Ralph Nader in 2000.) As an example, in Pew’s 2016 poll, Johnson and Stein combined for 18% of the votes among the youngest cohort. This year, the combined third-party share among 18-29s is only 12%, which is still much higher than any other demographic category, but obviously a big and consequential drop.
RELIGION
RELIGION |
CLINTON |
TRUMP |
2016 MARGIN |
BIDEN |
TRUMP |
2020 MARGIN |
CHANGE |
WHITE EVANGELICAL |
14 |
75 |
-61 |
17 |
78 |
-61 |
0 |
WHITE MAINLINE |
37 |
48 |
-11 |
43 |
53 |
-10 |
+1 |
BLACK PROTESTANT |
87 |
3 |
+84 |
90 |
9 |
+81 |
-3 |
CATHOLIC |
46 |
44 |
+2 |
51 |
44 |
+7 |
+5 |
UNAFFILIATED |
59 |
21 |
+38 |
71 |
22 |
+49 |
+11 |
Finally, let’s take a look at the religion crosstabs. Despite a few high-profile individual evangelical defections to Biden in recent months, the rank-and-file haven’t really budged. One place where Biden’s seen an appreciable gain, though, is with Catholics. That’s important because one place where Catholics are overrepresented is the populous swing states of the Midwest and Northeast.
Interestingly, that Catholic number (and the evangelical number) might also help us to make more sense about the Latino numbers we were discussing earlier. If you were to assume Latinos are monolithically Catholic, then seeing the two figuress going in opposite directions wouldn’t make much sense. However, an increasingly large share of Latinos in the U.S. are evangelical, so it’s possible that we’re seeing evangelical Latinos moving in the Republican direction, while Catholic Latinos are not.
The largest movement of any group, however, is among the “unaffiliated” (which is how Pew groups together both atheists and agnostics as well as people who simply answer “nothing in particular”). Notice that this is a very similar-sized move as the shift among voters aged 18-29, and that makes sense, since younger generations are much more likely to skew nonreligious than older generations. And as with the 18-29 bracket, note that the Trump number is unchanged; instead, a much larger number of them are committed to Biden this time rather than undecided or voting third party.