Nate Silver made virtually the exact argument that I’ve been making about the importance of the special election in the 6th District of Georgia that will take place tomorrow. First, he points out that most of the post-mortem takes you will read about this election on Wednesday are going to be “dumb” because the lessons we should learn will come from the fact the election is close (assuming the polls are right), and that knowing the winner won’t really add much of value. Then he says that sometimes dumb arguments can still matter if enough people believe and act on them. He’s absolutely correct about that, and a Democratic win might modify Republican behavior on a host of issues, including their health care bill and their treatment of the Russia investigation. A Democratic loss might have the opposite effect.

If Handel doesn’t win in a blowout, what we’ll want to examine is why this has happened:

Georgia 6 is a tough district to diagnose because its politics in presidential elections shifted a lot from 2012 to 2016. In 2012, the district went for Mitt Romney by 23 percentage points in an election that then-President Barack Obama won by 4 points nationally. That made it 27 points more Republican than the country as a whole. In 2016, by contrast, it chose Trump over Hillary Clinton by only 1.5 points in an election where Clinton won the popular vote by about 2 points nationally. Therefore, it was only 3 to 4 points more Republican than the national average.

Hillary Clinton managed to do something similar in many affluent, highly educated suburbs. Georgia’s 6th is just one of the starker examples, where she moved a district from being 27 points more Republican than the national average to being 3 or 4 points more Republican. What’s critical to understand, though, is that she did this (which was a key part of her plan), and she still lost because Trump moved rural areas even further in his direction.

Thanks to the design of the Electoral College and the Senate, political power in America has always been influenced almost as much by the geographical reach of the political parties as by their ability to attract the most raw votes. And over the past two decades, the left has suffered a very unfavorable geographic realignment.

One good way to see this is to look at the changes in county-level voting behavior in presidential elections. It wasn’t so long ago that a roughly equal number of counties supported the two main presidential contenders. In 1992, Democrat Bill Clinton nearly matched the performance of Republican George H. W. Bush, winning 1,519 counties to Bush’s 1,582. (Even then, the Republicans held an advantage; Bush won his counties with only 37 percent of the national popular vote.) The county totals were almost identical when Clinton defeated Bob Dole four years later.

That would turn out to be a high-water mark for the Democrats. Even while narrowly winning the popular vote in 2000, Al Gore carried only 659 counties to George W. Bush’s 2,397. John Kerry did even worse in the close 2004 election. But the worst performance since Walter Mondale’s 1984 shellacking came in 2016, when Hillary Clinton topped Donald Trump in only 489 of America’s 3,141 counties.

Overall, Donald Trump carried 220 counties that had voted for Obama in 2012, while Hillary took only seventeen that had gone for Mitt Romney. Even more significantly, Trump got a larger share of the vote than Romney in 2,728 counties. Clinton outperformed Obama in only 383.

That’s just one way of looking at how the country has been realigning and it’s only a piece of the puzzle that explains how Trump defied all the polls and won the Electoral College. But you’ll never come close to “getting” what happened if you don’t spend some time thinking about the fact that Trump did better than Romney in 2,728 counties and Clinton did better than Obama in only 383. The counties in Georgia’s 6th are among those 383 where Clinton outperformed Obama (or Trump underperformed Romney), and as a group these counties are where the Democrats are seeking to make their political comeback. They tend to be more highly populated counties which means that they have more political representation than the grouping of counties where Trump excelled. But the real problem for Democrats is that they are now doing so poorly in so many counties that they can’t field competitive candidates. As a result of this, the Democrats win a few seats in legislatures by huge margins and collect a respectable share of the popular vote, but have few options for winning narrow victories. Affluent suburbs provide their only remaining targets of real opportunity.

It makes sense to take wins where you can get them, and if a district has moved 23 or 24 points in your direction then the Democrats should certainly compete there. The problem is that there are districts all over the country where the movement has been in the opposite direction. Take a look at Pennsylvania:

When President Obama ran for reelection in 2012, he suffered a ten-point loss in the percentage of his 2008 two-party support [excluding third parties] in just two Pennsylvania counties (Midwestern Cambria and Elk). Clinton would suffer a 10-point reduction in two-party support compared to Obama’s 2012 performance in 23 of Pennsylvania’s 68 counties, and in 45 counties when compared to 2008.

The 2008 election was a high-water mark for the Democrats, but losing ten points off of that performance in 45 out of 68 counties helps explain why Clinton lost Pennsylvania despite outperforming Obama in the affluent well-educated suburbs.

It shouldn’t surprise you to learn that the Pennsylvania legislature looks far out of reach for the foreseeable future.

Now, some people think that this is unfixable. But I’d like to remind you that the Democrats did much better in these rural areas four years ago and especially eight years ago. Here’s an example:

One way to illustrate what happened is to look at two adjacent southwestern counties that border West Virginia.

In 2008, Obama essentially tied John McCain in Greene County, losing by just sixty votes. In 2012, Obama lost to Romney by 2,576 votes. But 2016 was a disaster: Clinton won a mere 29 percent of the Greene County vote, costing her 6,367 net votes. Trump picked up 14 percent of his statewide margin from a county that produced fewer than 16,000 total two-party presidential votes.

Just to the north, in more populous Washington County, the erosion was both less extreme and more consequential. Obama lost Washington County by 4,571 votes in 2008 and by 12,885 in 2012. In 2016, Clinton lost by 25,064, which was more than half of the statewide margin. These two lightly populated and heavily white working-class counties alone accounted for 71 percent of Trump’s margin of victory.

The rural tidal wave more than wiped out Clinton’s advantage in places like Chester County, in the Philly suburbs. Mitt Romney had carried the affluent and traditionally Republican county by 539 votes. Trump’s style, policies, and record of sexual assault weren’t expected to play with Romney Republicans, and they didn’t: Clinton won Chester by 25,568 votes. But that was essentially single-handedly neutralized by Washington County, which has a population less than half the size of Chester’s. Clinton won the big counties, but she lost the small counties so badly it didn’t matter. The state, along with the country, had realigned, but the realignment wasn’t an even trade.

Now, if the Democrats win the special election in Georgia tomorrow, that will show that they can take advantage of the half of this realignment that favors them, but it won’t do a thing to change the fact that the realignment isn’t an even trade. That’s the danger.