Yesterday I discussed the difference between polls that test registered vs. those that test likely voters. Well, the best way to highlight my point of how testing registered vs. likely voters affects a poll, is to do so using a single poll that asks both sets of respondents the same questions during the same period of time.
Well, the Franklin & Marshall Pennsylvania Senatorial Survey did just that.
They tested the preference of both registered and likely voters in the Senate race pitting Democrat Arlen Specter vs. Republican Pat Toomey.
Among registered voters, Specter won 33-30.
Among likely voters, Toomey won 44-34.
Quite a shift, eh?
Keep that in mind when reading a poll - testing registered vs. likely voters does make a difference, and fortunately for Republicans, testing of likely voters is much more often than not more accurate, but also highlights more Republican tendencies.