Note: I don’t intend for this to be an academic post as even I tend to glaze over when presented with equations and fancy-arsed academic language. However if readers would like to read more I’d recommend tracking down Luc Anselin’s work as a starting place and then graduating to others as the reading takes you. Just don’t expect me to run through complex formulae here; the truth is they’re intensely boring and nobody likes them. Ahem.

First, some boring pretext. You are about to see some maps with colours. The data used to create the maps are the US county results for the 2008, 2012 and 2016 US Presidential Elections from Dave Leip’s Election Atlas (thanks Dave!). The red counties are counties where there is high vote share for that candidate surrounded by counties (Queen contiguity) with equally high vote share for that candidate. The blue counties are counties where there is low vote share for that candidate surrounded by counties with equally low vote share for that candidate. There are a world of statistics I would be expected to describe here but I’m keeping it real here so academics will have to come to peace with it! REMEMBER: RED IS HIGH…….BLUE IS LOW*

Below are the maps for the performance of the Democrats since 2008.

Clustering of McCain vote share, 2008 (Moran’s I = 0.603)


Clustering of Romney vote share, 2012 (Moran’s I = 0.742)


Clustering of Trump vote share, 2016 (Moran’s I = 0.733)


Where the +/- change in vote share for the Republicans has clustered between 2008 and 2016


The final map shows where the change over that period has clustered. In that map the red areas show where an increase in vote share has clustered, whereas blue shows where a decreased has clustered. There are significant spatial patterns to both Republican support (unsurprising) and the change since 2008 (perhaps surprising).

It is perhaps interesting to note that Mitt Romney appears to have improved the Republicans performance in parts of Minnesota, Wisconsin and Michigan in 2012. Not totally unsurprising given Mr Romney’s associations with Michigan in particular. Donald Trump further improved the performance here of course, but perhaps one can see the seeds of recovery in GOP performance in 2012 in parts of the rust belt.

Mr Trump also continued the improvement in GOP performance in Appalachia, where the clustering of vote share in Appalachian counties in Pennsylvania, Ohio, West Virginia, Kentucky and Tennessee intensified further between 2012 and 2016.

GOP performance in Nevada, Utah and parts of Idaho weakened in 2016, probably due to Mr Trump’s unpopularity among Mormon voters in Idaho and Utah in particular.

*The light blue counties indicate Low vote share surrounded by High vote share. The pink counties indicate High vote share surrounded by Low vote share.