There are any number of policy disputes within the 2020 Democratic presidential nominating contest, from nuanced differences over expanding access to health care to sweeping debates over the best approach for resolving climate change. Those arguments, though, sit on top of another, more fundamental political question: What’s the best way to beat President Trump?

There are two schools of thought on that point. The first looks at the 2016 election and sees Democrats losing persuadable working-class white votes — and therefore the key Rust Belt states of Michigan, Pennsylvania and Wisconsin. The second looks at the same election and identifies a decline in turnout from President Barack Obama’s 2012 reelection in those states as the reason for Trump’s victory.

There’s validity to both ways of looking at what happened. Analysis completed by a team of researchers last year found that about 7 percent of 2012 Obama voters didn’t vote in 2016 and 9 percent voted for Trump. Nearly all those who went from Obama to Trump were white; the research suggests that 12 percent of white Obama 2012 voters supported the Republican four years later.

Addressing the first problem means trying to win votes back. Addressing the second problem means trying to reinvigorate voters who skipped the election three years ago.

There’s necessarily overlap here with the aforementioned policy disputes. After all, one might reasonably assume that voters wooed by Trump are less likely to embrace more liberal policy proposals. (Candidates such as Sen. Bernie Sanders (Vt.) reject this assumption, for what it’s worth.) Advocates of focusing on boosting turnout, meanwhile, suggest that innovation and clarity of ideology is the sort of thing that inspires people to make time to vote.

This debate gets simplified into a related question: To what extent do Democrats need to worry about winning over white working-class voters?

Bloomberg News’s Sahil Kapur put it this way in a tweet Sunday. (Lack of a college degree is a common way to identify working-class voters in polling.)

The first three numbers are correct, taken from exit polling conducted by Edison Research. The conclusion — about the inability to make “meaningful gains” — is disputed. Cook Political Report’s Dave Wasserman, for example, replied succinctly.

Kapur’s data can be read to suggest both that Democrats aren’t going to win working-class whites and, further, that it doesn’t matter. Democrats have lost the group by wide margins in elections they both won (2012 and the 2018 House races) and lost.

His assessment overlaps with analysis produced by Gallup this year. The venerable polling firm determined that since about 2006, partisan views of whites with or without college degrees have splintered. Whites without degrees have become consistently more Republican while those with degrees, after tracking more to the right for a while, started trending increasingly Democratic from about 2010 on.

If that’s the case, why bother trying to win over those working-class voters?

Wasserman’s point is that degree matters. In subsequent tweets, he notes that there’s a link between the states that Trump flipped in 2016 and the density of their working-class white populations.

We can visualize that pretty clearly. As the density of working-class whites among the over-18 population in a state increased, so did the margin by which it preferred Trump in 2016. It’s not a perfect correlation, but it’s a robust one.

On that graph, we’ve highlighted six states. Three are those that flipped to Trump three years ago: Michigan, Pennsylvania and Wisconsin. The other three are Arizona, Georgia and Texas, states that have consistently voted Republican — but which are also consistently mentioned as ones that might soon turn blue. As it turns out, those six states give us interesting insight into that fundamental Democratic debate.

Let’s start by looking at Kapur’s initial data. Here, we’ve plotted three elections: 2008 (which the Democrats won), 2016 (which they won in terms of actual votes, although not electoral votes) and the 2018 midterm. You can see the bulge Wasserman points to: In 2016, the gap between those top and bottom lines is much wider than the gaps in 2008 or 2018.

Notice something else, though: The density of the white working-class vote in 2018 (the gray line) was higher than in 2016. This is a function of the 2016 electorate being larger and more diverse, as presidential years generally are. Notice, too, that the density of the white working-class vote was also larger in 2008 — a function of our elections getting more diverse over time. The 2008 election that brought Obama to the White House had a then-historically low density of white voters — but the 2014 election in which Republicans did well was just as diverse.

That diversification is one reason some argue against the need for Democrats to appeal to white working-class voters. It is a group that has diminished as a percentage of the electorate even as the number of nonwhite voters — who generally vote much more heavily Democratic — increases. But while that shift is real, it’s not as significant as it may seem at times. Whites without a college degree still made up 40 percent of voters last year and still make up just under 45 percent of the over-18 population in the country.

Anyway. Back to Wasserman’s point.

Both Pennsylvania and Wisconsin had shifts from 2016 to 2018 that looked like the national shift, according to exit polls; the gap between support for Democrats and Republicans among working-class whites narrowed even as they made up more of the electorate. In Michigan, the gap with those voters widened. But in all three states, Democrats did much better in aggregated House voting than in the 2016 contest.

It’s really important to note that these are imperfect measures. Exit polling is useful as a guidepost but isn’t exactly gospel, particularly when looking at demographic subsets of an election. It’s also important to note that we are comparing presidential preferences in 2016 with gubernatorial preferences in 2018.

With that in mind, let’s compare the states above — the ones that flipped from blue to red in 2016 — to Arizona, Georgia and Texas.

You’ll notice first that the gaps in support among white working-class voters between the parties are much wider than they are among the first three states we looked at.

What really stands out here is Arizona: Democrats won more votes in House races, even though white working-class voters shifted more heavily to the Republican than in 2016 and made up more of the electorate.

Well, here we see the effects of the actual races. In Senate exit polling, white working-class voters cast their ballots about the same way they had in 2016. Gov. Doug Ducey romped over his Democratic challenger, while Kyrsten Sinema (D) edged out Republican Martha McSally in the Senate contest.

This state in particular seems to bolster Wasserman’s point: With a narrower gap among white working-class voters, Sinema actually won.

Another point not reflected in the graphs above is worth noting: In each of the six states except Pennsylvania and Wisconsin, the density of the nonwhite vote in 2018 was the same as in 2016. In those two states, nonwhite vote density was higher in 2018. The Democrats gained about the same margin in House races in those states as they did in Michigan, where nonwhite vote density was flat.

But it brings us back to the idea that turnout may also play a role.

The 2018 midterm election had lower turnout than the 2016 presidential contest, as always happens in the modern era. But compared with 2016, the number of votes cast for Republicans sank more than the number of votes cast for Democrats.

This could be a function of either Republicans turning out less heavily or of votes flipping from Republicans to Democrats.

What’s really interesting how actual vote totals compared in the six states we isolated.

In the three states that went from blue to red in 2016, vote totals for both parties were down in 2018, but Republican totals were down more. In Michigan and Pennsylvania, 2016 votes for the Republican (that is, Trump) look like a blip. In Wisconsin, the increase was in line with 2012 — but in 2012, a Wisconsinite was on the Republican ticket.

It looks like a surge in Republican votes coupled with a decline in Democratic ballots. Or, looking at the data another way, 2008 and 2012 look like a Democratic bubble. The 2004 results in each state were much closer to the 2016 results, particularly in Wisconsin, which Democrat John F. Kerry only barely won.

In each state, as Democratic vote totals dropped in 2012, Republican vote totals increased. 2016 looks like a surge in Republican votes, but clearly some of those voters were ones that four years earlier had voted Democratic. Boosting Democratic turnout in the states would have made a difference, but it seems likely given the narrow margins that stanching blue-to-red flips would as well.

For what it’s worth, the trend in the vote margin in all three states was broadly in line with the change from 2008 to 2012.

If 2016 was a function of voters flipping, these data fail to answer a key question: Which voters did so?

Now look at the other three states.

There, Republican vote totals have largely held steady, while Democratic totals have surged. In Arizona’s House races, the steady uptick in Democratic votes passed the Republican total — again, this was with nearly half of the votes in the state being cast by working-class whites who overwhelmingly backed the Republican governor. More people statewide voted for Democrats in the 2018 House races than voted for Hillary Clinton.

These six states are seeing diverging population trends. The ones that people think may soon turn blue are growing quickly — which is one reason people think they may turn blue.

Those new residents are often moving to cities, helping shift the states’ politics. As we noted last month, there were 15 states in 2016 where at least 80 percent of the vote came from urban counties. Clinton won 11 of them. The four she lost were Florida and Pennsylvania (which she lost by about one percentage point each) — and Arizona and Texas.

We can take from Arizona’s 2018 election that there is certainly a need for Democrats to ensure that their deficits with white working-class voters don’t get too wide. From the Rust Belt, a warning about how the aberration may have been 2008 and 2012, not 2016, in those heavily white-working-class states.

There’s an overarching issue that’s worth pointing out, a third layer to the discussion: Will resolving this issue matter in 2020?

2016 was historically close. Exceptionally close. We spend an enormous amount of time reconsidering it simply because it’s the most recent election and the one that brought the likely Republican nominee in 2020 to power. But it’s very possible that these analyses of the thinly sliced margins in 2020 simply won’t matter much. That the contest will be a lot more like a Doug Ducey-style romp than a Kyrsten Sinema-type nail-biter.

This debate, in other words, is largely about a contest that is in the past and that may not be repeated in 2020. The white working-class vote may not matter simply because the election ends up looking a lot more like 2008 than 2020.

In which case it’s 2016 that’s the outlier, not necessarily something to be studied to learn new details about where the country is headed.