Wait, I thought that Rasmussen is "the best".
I always hear from conservatives that Rasmussen is the most accurate because of "likely voters", and that their constant outlier results are because they're the only ones who are right.
I guess that's only "true" when you like the results...
[MENTION=28109]Amelia[/MENTION]
Yes, in relation to other polls, Rasmussen (or better put, not-Rasmussen, since Scott Rasmussen has left the firm) is indeed an outlier.
But the approval polls are three day rolling polls. Not all firms do this. Some, but not all.
Rasmussen has a strong history of being the outlier.
In the 2008 presidential election, Rasmussen was consistently off to the RIGHT by about 4 points across the board in polling margins.
In 2010, it was the same.
In 2012, it got worse:
Statistikhengst's ELECTORAL POLITICS - 2013 and beyond: The moment of truth: how did the pollsters do?
Rasmussen put out end polls in 21 states in 2012, 12 of them were the big battleground states:
FL, OH, VA, NC, WI, IA, NV, NH, PA, MI, MN, CO.
Of those 12 battlegrounds, Rasmussen missed the call in 6 of them.
Rasmussens mathematical bias (average) of all 21 states was:
R +2.81
Of those 21 states, Ras was off to the left in polling in 5 states, it nailed the marigin in 1 state, and it was off to the Right in 15 states.
But of the 12 battleground states, it was
R +4.50
Of those twelve states, Ras was off to the Left in polling in one state (NM), it nailed the margin in one state (PA) and it was off to the right in 10 states.
Of the 7 major pollsters, there is no doubt that Rasmussen was by far the worst and most unreliable in the state polling in that year.
By mathematical bias, I mean the average of the polls in the given set, the average of their distance from the actual results.
For instance, if a pollster does three polls:
poll 1 is 4 points to the left
poll 2 is 5 points to the left
poll 3 is 1 point to the left
4+5+1= 10 / 3 = average of 3.34 points to the Left.
It's pure, simple math. And all explained at the link above.
It was so unbelievably comical in
2010: I used to see a Ras poll, and then I would mentally move the margin
at least 4 points to the Left, and bingo, that ended up being the end result:
2010 WA Senate: Ras said Rossi +1. Murray won by 5.
2010 NV Senate: Ras said Angle +1. Reid won by 5.
And the list goes on and on and on.
The special Senatorial in 2010, early in the year, for Scott Brown, that poll Rasmussen did get right.
Rasmussen also has a very strong tendency to suddenly "find" lots of undecided voters in Blue leaning states and in states where we pretty much know that the Democrat will win and come well over 50%, Ras will have that DEM candidate under 50%.
Ras tends to find lots a lots LESS undecided voters in red leaning states in poll, and I can prove it, for I have collected every single ras poll from 2008, 2010 and 2012. Well, actually, I have collected every single poll out there.
Many pollsters open up their crosstabs for others to see, including all internals. Ras does not.
Hope that information helps. This is Rasmussens exact track record, esp. in end-polling. It is a record that Rasmussen cannot hide from, because the topline data was made public.