President Trump’s promise to voters during the 2016 election was that he would rewind the U.S. economy. What had made America great, he would argue, was domestic manufacturing and resource extraction — mining, oil drilling, that sort of thing. There was a nostalgia to the pitch that appealed to his target audience, and there was some internal logic to it: The postwar boom involved a lot more steel production than, say, Internet search engines, for obvious reasons.

The problem with the argument was that it wasn’t very practical. Fracking kicked off a big expansion in oil and gas production under President Barack Obama, but only about 165,000 people are employed in extraction. Coal mining adds another 53,000, an increase of only about 2,000 since Trump took office. The service industry, on the other hand, is massive. By comparison, 227,000 people work in retail trade … in the Seattle area alone.

Much of the economic growth in the United States in recent years has been in urban areas. This is in large part why Trump was focused on the issue in the first place; instead of those urban coastal elites seeing all the gains, he would bring gains back to the heartland.

New data from the Census Bureau, though, suggest that, even under Trump, it’s bluer areas of the country that are seeing bigger gains.

There are several reasons for this, including the increasing centralization of technology jobs in certain regions, as the Brookings Institution reported this month. Part of it, too, is that areas that once relied on manufacturing haven’t evolved their economies enough to prevent younger workers from migrating elsewhere, as Well Fargo Securities economist Mark Vitner told the Associated Press. The AP was reporting on the new census data, which found “household income grew the most in tech and entertainment centers like Austin, Texas; Nashville, Tennessee; and large chunks of the West Coast.”

If we break out that data by congressional district and overlay 2016 voting preference, clear patterns emerge. Incomes in districts that voted for Hillary Clinton in 2016 or more narrowly preferred Trump tend to range across a number of median incomes. Districts that were more heavily supportive of Trump are clustered in lower income ranges. (These figures are not adjusted for inflation.)

As noted above, this correlates to urban-rural splits. Across the board, more rural congressional districts see lower household incomes than more urban ones, regardless of politics. (Analysis of density comes from CityLab’s density index.)

That cities tend to be more heavily Democratic — there are 33 “pure urban” districts that voted for Clinton and one (Staten Island) that voted for Trump — influences the income data. The effect, though, is the same: Blue areas are seeing more income growth.

Again, areas are seeing gains regardless of politics. Incomes increased from 2014 to 2016 and from 2016 to 2018.

The jump in more heavily Democratic areas was larger, with the most heavily pro-Clinton areas seeing mean household income increases of over $15,000. The most heavily pro-Trump areas saw increases of less than $10,000.

As a percentage, the increases since 2016 in more-Democratic areas were also higher. In the most heavily pro-Trump areas, there’s not much difference in the change in household incomes between 2014 and 2016 and between 2016 and 2018. In less polarized areas, the recent increase has been sharper.

The extent to which Trump can shape the economy was always more limited than he suggested. He has enacted some policies, such as tariffs, that have had an effect opposite to the one he promised. (Thanks in part to those tariffs, employment growth in the Midwest is trailing other areas of the country.)

The implicit promise of Trump’s campaign, though, was that red America would thrive at blue America’s expense. Under Trump, blue America is doing just fine.