As we get closer and closer to November, the number of polls proliferate, and everybody is spitting out statistical and anecdotal analysis, it is important to revisit how the lessons from 2016 exit polls should be kept in mind when reading polls or political analysis.

First-off, the exit polls from 2016 paint a fairly inaccurate picture of what transpired that November. While country data saw many ancestrally blue counties shift blue, many similarly red counties shifted blue. Yet, exit polls suggested Trump managed to win over the college educated white voters who dominate in these areas. The problem is this flies in the face of what we actually intuitively assume. Yet it dominates political analysis.

Take Politico’s recent series on eight battleground states (MI, NC, AZ, FL, WI, PA, MN, GA). In particular, their analysis of Michigan where the article concludes with the following:

“Add it all up — Trump’s ceiling with the white working-class, the exodus of suburbanites away from the GOP, and a near-certain turnout jump among Black voters — and the state of the race in Michigan looks grim for Trump. The three dynamics that all broke his way in 2016 now all appear to be breaking against him, and with voters in the state soon receiving their absentee ballots, the president’s allies on the ground expect he could concede Michigan sooner rather than later.”

Yet, the article relies on exit polls for its analysis and conclusion. Yes, Trump held down his losses in Oakland County to eight points and black turnout was down. However, considering how well Trump ran among non-college educated whites it is likely these voters helped him hold down his margin in Oakland. Two years later without him on the ballot and some of these voters, particularly in Macomb, either stayed home or voted blue at the local level. But, notably, black turnout in 2018 was not exponentially higher (ie. no Democrat has replicated what Obama did with his brothers and sisters). So, an assumption Trump should pull out because of these factors is the kind of analysis which led to 2016 surprising.

The Pew Research Center and the Center for American Progress have produced methodologically sophisticated surveys of the electorate that sharply contradict 2016 exit polls. For example, contrary to the popular wisdom Democrats are increasingly the party of college educated whites, in November of that year Pew found 33 percent of Democratic voters and leaners actually were whites without college degrees. Only 26 percent of whites with college degrees identified as Democrats. The rest identified as non-white with 28 percent not having degrees and only 12 percent with.

In other words, the analysis the party is one of minorities and college educated whites is a bit overstated. To date, the party has largely campaigned on appealing to its urban, college educated base while throwing enough crumbs to its diverse non-college educated base to get them to turn out. Polling suggests this won’t harm the party in November. But, then again, polls were wrong once already.

In the detailed analysis of the 2016 vote, Pew found that 44 percent, or 60.1 million out of a total of 136.7 million votes cast on Nov. 8, 2016 were cast by whites without college degrees — demographic shorthand for the white working class. Exit polls suggested the white working class only cast 34 percent of all votes or 46.5 million out of 136.67 million cast. This means the white working class cast 13.5 million more votes than exit polls suggest.

Hillary Clinton won 28 percent of white working-class votes, according to Pew, less than Obama’s 36 percent in 2012. Still, a quarter of her total vote of 65.85 million — that is, 16.8 million votes — came from the white working class. Admittedly, some of these voters are shifting to Trump but me be less likely to tell pollsters than Romney to Clinton to Biden voters.

At the same time, Pew found that whites with college degrees made up 30 percent of the total electorate, not the 37 percent reported in the exit polls. In other words, Pew found that white working-class voters outnumbered white college voters among all voters, while the exit polls reported just the opposite.

This skews analysis and paints a fairly rosy picture for Democrats going forward. Afterall, if white college educated voters are trending blue and more likely to either be picked up in polls or respond, their preferred party will benefit. At least in polls. More importantly though, these numbers have powerful ramifications for both Democrats and Republicans preparing for November and beyond.

For Democrats, the numbers suggest there remains an equal downside to benefit ratio for practicing identity politics. Sure, you might win diversifying suburbs but it will increase the reddening trend of the Midwest. For example, per CAP’s Robb Griffin and John Halpin, whites without degrees outnumbered those with degrees 45 percent to 29 percent.

Remember, exit polls taken on Nov. 8 had Clinton failing to carry college-educated whites, losing them to Trump (49-45). Pew found that she did in fact win these voters, decisively carrying white college grads, 55-38. This has particular ramifications for states like Michigan and Pennsylvania, even if Trump did better there than nationally.

According to the November exit polls, half of the entire 2016 electorate of all races had college degrees; Pew found that such well-educated voters were a much smaller 37 percent. Admittedly, exit pollster Edison Research does not deny their methodology – interviewing voters as they leave the polling place – is imperfect. A criticism of exit polls since their inception is only those who want to talk about their vote participate, skewing the results.

It is true exit polls do try their best to weight the results of even those participating. But, all they can do is weight according to age, race and gender and then approximate (guess) what the electorate looks like beyond that. In many ways, taking the same approach as traditional polling.

To many pollsters and academics, the issues with exit polls, and polls in general, are well known. But, to the media and journalists, who report on every poll like it is a hard and fast number set in stone for all eternity (MoE’s single snapshot in time don’t exist), these issues are rarely discussed or considered. If one were to consider this, they might think the polls have a statistical bias to Democrats due to who responds to polls (college educated, urban voters).

Further, another issue with exit polls is they only provide a representation of the actual voting electorate and not eligible voting population. Because of these issues, most academics and political scientists look to other sources to study voting behavior.

Heck, even Nate Cohn argued that the exit polls overestimated “Trump’s support among well-educated white voters” and that……

“there is no question that the exit polls underestimate the number of white working-class voters, especially older ones, by a considerable amount.”

Cohn observed that errors result from “an odd methodological quirk in exit polling” that “winds up biasing the rest of the survey because the exit polls are weighted to match the actual result of a far less educated country.” The net effect is exit polls tend to assume college educated voters are more Republican than they really are while overestimating the Democratic lean of non college voters and the young.

So, what does this all mean for November? Well, for Democrats it means they cannot forget their blue-collar white and diverse base. Preventing further erosion among this group is crucial if they want to retain states like Minnesota beyond just November.

For Republicans, while they may comfort themselves they can win national elections with what they have – albeit by narrow margins, they cannot afford to suffer continued losses among college-educated white voters, especially college-educated women. Pew found by 2017, a year into the Trump presidency:

“Voters who have completed college make up a third of all registered voters. And a majority — 58 percent — of all voters with at least a four-year college degree now identifies as Democrats or leans Democratic, the highest share dating back to 1992. Just 36 percent affiliate with the Republican Party or lean toward the G.O.P.”

For both parties, the danger has become more acute. For the GOP, the share of women identifying as Democratic or leaning that way is at its highest point since 1992. For Democrats, voters with no college experience have moved towards the GOP with a whopping 47 percent identifying as Republican and many voting that way consistently.

The problem with analysis, even political analysis from places which should know better, is exit polls set the stage for much political analysis. For a long time after the election, there is little to refute the narrative which leads to flawed analysis and a continued questioning of our institutions and political leaders.

It’s hard to argue Democrats don’t know they need to do better with non college educated white voters. Arguably, it is why Joe Biden was selected as the nominee and why he has played up his Scranton roots. Yet, on the issues that matter to these voters, immigration in particular, Biden has been lacking. His pledges of not raising taxes benefit the wealthy as much as the blue-collar worker. His refusal to go after China the way Trump has also damages his brand among these voters.

Putting it bluntly, if Clinton had performed as well among non college educated voters as Obama she would have carried, with robust margins, the states of Michigan, Wisconsin, Pennsylvania, and Iowa, as well as Florida and Ohio. In fact, if Clinton could simply have reduced the shift toward Donald Trump among these voters by one-quarter, she would have won.

In Alabama’s 2017 special Senate election, Doug Jones, the winning Democrat, would have lost if he had not made substantial inroads with the white working class, Teixeira argued:

“Without the hefty swing among the white non-college population, particularly women, there is no way Jones would have won the state, or even come close.”

It is likely why Teixeira concluded, and I agree:

“There is no way around it — if Democrats hope to be competitive in Ohio and similar states in 2020, they must do the hard thing: find a way to reach hearts and minds among white non-college voters.”

But, let’s go back to the write, William Galston, who I think summed up on the Brookings website, presciently, in June 2016, what was coming. I will quote him at some length, because in my opinion no one captured/s the situation as much as he does. Perhaps it suggests something coming in two months?

Most working-class whites have incomes below $50,000; most whites with BAs or more have incomes above $50,000. Most working-class whites rate their financial circumstances as only fair or poor; most college educated whites rate their financial circumstances as good or excellent. Fifty-four percent of working-class whites think of themselves as working class or lower class, compared to only 18 percent of better-educated whites ….

In many respects, these two groups of white voters see the world very differently. For example, 54 percent of college-educated whites think that America’s culture and way of life have improved since the 1950s; 62 percent of white working-class Americans think that it has changed for the worse. Sixty-eight percent of working-class whites, but only 47 percent of college-educated whites, believe that the American way of life needs to be protected against foreign influences. Sixty-six percent of working-class whites, but only 43 percent of college-educated whites, say that discrimination against whites has become as big a problem as discrimination against blacks and other minorities. In a similar vein, 62 percent of working-class whites believe that discrimination against Christians has become as big a problem as discrimination against other groups, a proposition only 38 percent of college educated whites endorse.

This brings us to the issue of immigration. By a margin of 52 to 35 percent, college-educated whites affirm that today’s immigrants strengthen our country through their talent and hard work. Conversely, 61 percent of white working-class voters say that immigrants weaken us by taking jobs, housing, and health care. Seventy-one percent of working-class whites think that immigrants mostly hurt the economy by driving down wages, a belief endorsed by only 44 percent of college-educated whites. Fifty-nine percent of working-class whites believe that we should make a serious effort to deport all illegal immigrants back to their home countries; only 33 percent of college-educated whites agree. Fifty-five percent of working-class whites think we should build a wall along our border with Mexico, while 61 percent of whites with BAs or more think we should not. Majorities of working-class whites believe that we should make the entry of Syrian refugees into the United States illegal and temporarily ban the entrance of non-American Muslims into our country; about two-thirds of college-educated whites oppose each of these proposals.

Opinions on trade follow a similar pattern. By a narrow margin of 48 to 46 percent, college-educated whites endorse the view that trade agreements are mostly helpful to the United States because they open up overseas markets while 62 percent of working-class whites believe that they are harmful because they send jobs overseas and drive down wages.

It is understandable that working-class whites are more worried that they or their families will become victims of violent crime than are whites with more education. After all, they are more likely to live in neighborhoods with higher levels of social disorder and criminal behavior. It is harder to explain why they are also much more likely to believe that their families will fall victim to terrorism. To be sure, homegrown terrorist massacres of recent years have driven home the message that it can happen to anyone, anywhere. We still need to explain why working-class whites have interpreted this message in more personal terms.

The most plausible interpretation is that working-class whites are experiencing a pervasive sense of vulnerability. On every front — economic, cultural, personal security — they feel threatened and beleaguered. They seek protection against all the forces they perceive as hostile to their cherished way of life — foreign people, foreign goods, foreign ideas, aided and abetted by a government they no longer believe cares about them. Perhaps this is why fully 60 percent of them are willing to endorse a proposition that in previous periods would be viewed as extreme: the country has gotten so far off track that we need a leader who is prepared to break some rules if that is what it takes to set things right.

The bottom line, as the 2016 election amply demonstrated, is that if the Democratic Party does not take the bull by the horns, someone else will. It also suggests, the 2020 election might not go to form either.