This model uses state by state economic data. What is interesting about this model is the claim that it has worked 9 elections in a row and even accurately predicted the 2000 result where Gore wins popular and Bush wins electoral. The model itself is a new one, the 9 for 9 claim comes from the professors plugging in the state by state economic data on record in the election years all the way back to 1980 into the mathematical formula to see how close it was to the actual result (it doesnt claim to have been around back then, its just a retroactive application of the formula using the data from back then).
We'll see how close they come. There are a couple other models out there which project Romney at around 53% of the two party share (The famous Bread and Peace Model among others). Ofcourse, there are several models including Allan Lichtman's Keys to the Presidency, which predict re-election of the President.
I wouldnt be surprised by any result, be it a slight Romney win, slight Obama win, or a win by any of the candidates by 3-4 points.