Gallup Polling released a bombshell today when it quietly applied a likely voter screen to its generic ballot polling. As followers of this website know, only likely voter samples have any real predictive value for the midterm elections in the House of Representatives according to Gallup itself. Last week, Gallup showed the two parties tied in the generic ballot using a registered voter sample. Today they applied a likely voter screen to both last week's results and this week's and suddenly the Republicans have a 13 point lead in one scenario and 16 in another one.
Their analysis is very instructive. They start with a sample of 3,037 adults for the last two weeks. Of these, 2,764 are registered to vote, but only 1,882 are considered likely to vote. As we have repeatedly told you, adult polls do not be mean beans in politics and should be ignored. If you are not registered you cannot vote so you have no influence on an election. But not all registered voters actually vote so registered voter polls only tell you what would happen if everyone actually turned out which never happens. Turnout varies from election to election being much higher in presidential elections and much lower in midterm elections.
Republicans turn out better in all national elections and even more so in midterm elections so registered voter samples under-represent votes for Republicans. Only likely voter samples correct for this bias. This is illustrated by the registered voters in this sample going for the Republicans by a 3 point margin. However, the likely voters show a whopping 13 or 16 point spread depending on the overall turnout. As you can see this is a huge difference and accounts for the so-called enthusiasm gap where the Republicans are more motivated to turnout in 2010 by about 20 points than the Democrats.
So what are the practical results of this huge gap? We have provided a table that allows you to convert the generic ballot spread in likely voter samples to actual seats in the House of Representatives. A 13 point spread blows of the end of the table because the Republicans have never gotten better than a 7 point spread in any midterm election since 1950 and trying to extend the table out that far gets to be mathematically imprecise. Still even if Gallup's lower estimate of 13 points is correct, we are looking at the greatest blowout election for the Republicans in the last 80 years.
We can only estimate what the House would look like after such an election by extending equations out beyond any known data points which we have done in our ongoing series, Who is Winning the House? As you can see, a 13 point spread translates to the Republicans gaining 92 new seats while losing 2 seats of their own. That net gain of 90 seats would produce a 268 to 167 House for the Republicans and a 101 seat margin. The 16 point spread blows off the end of even the extended table and takes off into the wild blue yonder.
We find it hard to believe that kind of spread especially when Rasmussen after being as high as a 12 point spread is suddenly showing only a 3 point spread. Still it does tend to validate our own prediction of a 256 to 179 House for the Republicans with a net gain of 78 seats and a 77 seat margin.