Employment Going South. Literally: From 1990 to 2015, the US economy as a whole increased employment by about 30 million (30,056,664 according to the BLS). Employment in 1990 was 118,900,000, so that meant that there was a roughly 25% increase in employment over that 25 year period, with a little more than 1 million jobs added per year, on average.

I’m not going to surprise anyone by saying that these 30 million jobs were not spread equally across the entire US. What I didn’t have a good sense of, personally, was how disparate the change in employment was across the US. This post is just some documentation about the absolute size of the change in job distribution across the US. …

[analysis, graphs]

… Again, I’m not claiming that this is some kind of revelation here. The movement of population, and hence jobs, from the Northeast/Midwest to the Sunbelt is well known. What I found interesting was putting some tangible numbers of the shift. The little counter-factual I’m doing here is not very rigorous; there is no particular reason to believe that the 1990 distribution is the “right” distribution of jobs to compare against. But the 1990 distribution does have the feature of being prior to NAFTA and prior to China’s accession to the WTO, both of which are at times cited as sources of manufacturing job losses in the upper Midwest and Northeast.

The scale of the relative job changes, though, indicates that more of the losses have to do with free trade within the US than free trade outside of the US. The areas with relative decline lost 13 million jobs compared to the 1990 distribution of jobs. In total the US shed 6 million manufacturing jobs from 1990 to 2015 (18 million to 12 million, roughly). So this relative decline cannot possibly be a function only of manufacturing and international trade in manufactured goods. There is just too much relative movement out of the declining counties to attribute to this. This is a sloppy way of thinking about how this would work counter-factually (I’m ignoring spillovers entirely), but if you magically added 6 million extra jobs to those counties in relative decline, they would still be in relative decline compared to the Sunbelt in terms of jobs. They’d have 84.4 million jobs (as opposed to 78.4 million), but you’d expect them to have 95.6 million based on the 1990 distribution, so they would still be 11.2 million jobs off the pace.

The breadth of relative loss, though, seems striking, and is the one thing I did not appreciate prior to looking at this data. 909 counties lost jobs in absolute terms, which is 29% of all counties. Another 1,279 counties, 41% of all counties, gained jobs in absolute terms buy lost in relative terms. 944 counties, 30%, gained in relative terms. Just as many counties gained in relative terms as lost in absolute terms. The winning locations – Houston, Dallas, Atlanta, Miami, Phoenix, Denver, Vegas – won big, but the losing was spread across a wide area.

More jobs in 2015 are still located in places in relative decline (78.4 million) than are in places in relative ascent (70.5 million). Of those 78.4 million, 18 million (or about 12% of jobs) are in counties that experienced absolute job losses over the last 25 years. Most jobs are still in places that look to be losing out to Sunbelt cities over time. To the extent that your local economy plays a role in forming your opinions, this seems relevant, although I am going to stop now before I try to do any amateur political science or sociology.