As most people familiar with the literature of the subject will agree, one actual reason for abiding faith in a last-minute Republican “wave” this year is that early polls tend to be of registered voters while late polls focus on likely voters, which is a more Republican universe, particularly in a midterm with a Democrat in the White House. So the RV to LV “switchover” should produce a GOP “surge,” right?
Yes, in theory, but (1) some pollster use LV screens all along, (2) some forecasts bake a switchover factor right into the cake, and (3) the RV/LV gap is by no means uniform.
As an example, check out this notation from Nate Silver yesterday towards the end of a long post reporting some good polling news for Republicans:
A mild piece of good news for the Democrats is that the turnout gap may not be as large as it was in 2010. Both YouGov and NBC/Marist released results among both registered voters and likely voters in each of the states they polled — and on average, their likely-voter models showed the Republican candidate doing a net of about 2.5 percentage points better than in the registered-voter version of their surveys. That’s in line with the historical average gap between registered and likely voters in midterm years — rather than the 6-point gap that persisted throughout 2010.
So again, all the comparisons you’re seeing of the 2010 “wave” may well exaggerate the relevance of that very strange year.