Data from G. Elliott Morris’s model indicates the president is now underwater even in his final stronghold, Greater Appalachia
By Colin Woodard
Just how unpopular have Donald Trump and his policies become, now that he’s mired the country in a war nobody understands that’s bringing economic pain worldwide? According to the latest, highly detailed estimates from G. Elliott Morris – the data journalist who headed FiveThirtyEight until ABC shut it down last year – the president has messed up so badly he’s now underwater even in his ultimate regional stronghold, Greater Appalachia.
This winter we posted on Morris’s local Trump approval estimates based on his January 14-20 national poll, which indicated he’d hemorrhaged support everywhere outside Greater Appalachia and the eastern parts of the Far West compared to the 2024 election. He was already underwater across most of the Deep South and throughout the Lower Rio Grande Valley in El Norte, regions he won by wide margins in the last election.
But Morris’s updated approval estimates, based on his Strength in Numbers/Verasight poll fielded May 18th and 19th, make those numbers look rosy. Compare these maps, side by side, which display Morris’ local estimates as he reported on his substack, overlayed with the American Nations model boundaries. We don’t have the underlying data for his January estimates, but eyeballing it, it’s hard to find any location where the president hasn’t lost support in the past four months. From the coal fields of eastern Kentucky to the El Paso desert, Trump’s supporters are bleeding away.
These estimates are based on a national poll which shows Trump’s federation-wide support has fallen from minus 18 to minus 23 percent since January. Morris then cross-referenced those results with detailed community profiles ge has built from the results of surveys of 12,000 individuals and precinct-level demographic data. He then used statistical techniques to create inferential estimates of Trump’s approval in each of the country’s 2400 Public Use Microdata Areas, or PUMAs, an obscure Census Bureau unit containing roughly 100,000 adults. (The maps you see above depict PUMAs, not counties.) You can see his estimates for each PUMA in the interactive map at his site. (Zoom in to see the PUMAs rather than just the statewide estimates.)
We were able to obtain Morris’s PUMA-level approval estimates and convert them to the American Nations regions, weighing the results of each PUMA by its population. From that we were able to create a rough estimate of Trump’s current support in each region. Unsurprisingly, he has very few supporters in communitarian regions, ranging from under 35% approval in the Midlands to just 22.3% in Hawaii, part of the Greater Polynesian cultural space. Much more remarkable is how little support he now has in the individualistic regions that have long formed the core of his public and congressional support, Deep South (38.1%), Far West (38.5%) and Greater Appalachia (42.8%). As I noted with the earlier poll, I wouldn’t put any stock in Morris’ First Nation estimate, as the U.S. portion of that Arctic-spanning indgenous culture region is lumped into a single PUMA — “Subsistence Alaska” — along with a couple Far Western county equivalents. (The American Nations model uses county units.)
For comparison, these are the results of the 2024 presidential election by regional culture (left) and by county (right). Notice Trump won the Far West and Deep South by double digits, and Greater Appalachia and the New France enclave on southern Louisiana by more than twenty. If Morris’s estimates are even half right, they are a devastating rebuke of Trump and Trumpism.
As I noted in February, you should take all of this with a grain of salt. These are inferential estimates based on a statistical model and driven by a national poll of less than 1600 Americans – a solid sample for gauging national-level public opinion, but by no means definitive for sub-state analysis. The margin of error for most of the individual microdata areas is between 7 and 12 percent, theoretically enough to switch a few of them from positive to negative territory. Morris’s model is based on assumptions that may not turn out to be true. And, most critically for election forecasting, a lot of people may disapprove of Trump, but might vote for him anyway because they dislike his opponents even more.
But Morris’ estimates remain consistent with what we’ve seen in recent months whenever voters have gone to the polls, with Democrats greatly over-performing their 2024 performance across the country, and polling aggregators finding Trump reaching the lowest levels of support ever, in either of his presidential terms. At a time when the Alaska and Ohio US Senate races are competitive and pundits are once again talking about the possibility of Democrats turning Texas blue, these local approval estimates pass the straight face test.
Thanks, as always, to our partners at the culturally-minded geospatial consultancy, Motivf, where Tova Perlman wrangled this data and Aimee Trehey created the map at the top of this post.
— Colin Woodard, author of Nations Apart: How Clashing Regional Cultures Shattered America and American Nations: A History of the Eleven Rival Regional Cultures of North America, is director of Nationhood Lab.