The formula looks at factors such as population size and density of the countries people are leaving as well as those they are entering, and the distance between those places.
There's nothing earth shattering about considering the above variables when predicting migration. How those variables are weighted is another story, but you get a basic idea of what is important. Proximity definitely matters. A good model needs to control for population size and distance between origination and destination.
A lot of the error in any model can be explained with chain or network migration. I started out this blog mostly interested in this exception to the migration rule and I'm genuinely excited to find out how the new model accounts for distance. However, I'm now looking at the proximity advantage the Rust Belt possesses. When people leave shrinking cities, they aren't likely to travel very far.