Donald Trump won the presidency because he flipped a number of states that Mitt Romney lost into the Republican column. We mostly talk about the Rust Belt/Midwestern states -- Michigan, Pennsylvania and Wisconsin -- because all three voted for Barack Obama in 2012 and all three preferred Trump by a little over 100,000 votes, combined. But Trump flipped other states, too: Iowa, Florida and Ohio, the latter two of which were essential to his electoral college victory.

That victory, as we have heard repeatedly since Nov. 8, didn't extend to the popular vote. As it stands, Hillary Clinton leads Donald Trump in the popular vote by a margin of nearly 2.4 million votes. That and $2 will buy Clinton a weekend copy of The Washington Post at her local newsstand.

There are some interesting patterns buried within that split, though. Clinton's margin over Trump in California was one million votes larger than Obama's over Romney four years ago, for example. Trump's margin over her in Texas, meanwhile, was nearly 450,000 votes smaller. His margin in Arizona was over 100,000 votes smaller than Romney's over Obama. In the five states where the most votes were cast, Clinton outperformed Obama's margin over Romney by more than 400,000 votes, even while her margin nationally was 2.6 million votes narrower than Obama's.

The point has been made repeatedly that Clinton got more votes where she didn't need them. Her million-vote-larger popular vote spread in California earned her 55 electoral votes, same as Obama.

In fact, the pattern nationally was that states where the population grew more since 2012 also were more likely to see the margin between the Democrat and the Republican shift to the left. In other words, Clinton did better against Trump than Obama did against Romney as the percentage of the population increase went up.

(This isn't raw numbers of residents, mind you. It's increase in population.)

Some outliers to that (fairly loose) rule. Utah is relatively fast-growing, thanks to its young population. It moved sharply to the left out of disdain for Trump (and affection for Romney). North Dakota has been the fastest growing state in the country over the last decade thanks to the fracking boom, so it floats at the upper right in the chart above. West Virginia actually shed population between 2011 and 2015 (the years used for this estimation).

If we dig down a little, we find that the same correlation exists at the county level, though not quite as robustly. The less a county grew between 2011 and 2015, the more likely Trump was to see an improvement in his margin over Clinton relative to how Romney did against Obama.

Why? For one thing, the counties that grew the most were also largely those in which most of the inhabitants lived in urban areas (per the Census Bureau's 2010 numbers).

(Incidentally, North Dakota blows out the scale on those latter two graphs, thanks to the huge increase in places like McKenzie County, in the heart of fracking country. The solid diagonal lines indicating the trend include those data.)

Counties where a fifth of the population or less lived in urban areas in 2010 saw the presidential margin shift 15.2 points to the Republican in 2016 versus 2012. Counties where 80 percent or more lived in cities moved 0.6 points to the Democrat.

The vote shifts we started with are skewed a bit by the really large cities: the margin for Clinton in Los Angeles County was 260,000 more than the margin for Obama while the same shift in Cook County (Chicago) was 165,000. But the loose trajectory is this: People who moved to urban areas voted more heavily for the Democrat than in 2012. People everywhere else voted more for the Republican.

If that continues over the long term -- people moving to cities and voting Democratic -- it's not a good sign for the GOP, this year's surprise result notwithstanding.