Here is an interesting viewpoint on the midterm congressional polling...

This morning I saw that RCP replaced the ABC/WaPo "registered voter" poll results of +1% for GOP with the "likely voter" model result of +5% for the GOP.  That was a four point swing towards the Republicans just by looking at the more historically accurate sample. Makes you wonder if there would be a similar swing of RV polls were switched to Likely voter samples. So I just broke it down.

Overall voter congressional polling from RCP

Polling Data

Poll	Date	Sample	Democrats (D)	Republicans (R)	Spread
RCP Average	9/6 - 9/22	--	45.3	44.9	Democrats +0.4
ABC News/Wash Post	9/18 - 9/21	LV	46	51	Republicans +5
Emerson	9/20 - 9/21	1368 LV	45	45	Tie
Rasmussen Reports	9/18 - 9/22	2500 LV	42	44	Republicans +2
Economist/YouGov	9/17 - 9/20	1318 RV	45	40	Democrats +5
Politico/Morn Consult	9/16 - 9/18	2005 RV	46	41	Democrats +5
NBC News	9/9 - 9/13	1000 RV	46	46	Tie
FOX News	9/9 - 9/12	1201 RV	44	41	Democrats +3
NY Times/Siena	9/6 - 9/14	1399 RV	46	44	Democrats +2
Trafalgar Group (R)	9/7 - 9/9	1081 LV	42	48	Republicans +6
Harvard-Harris	9/7 - 9/8	1885 RV	51	49	Democrats +2

Registered voter congressional polling

Polling Data

Poll	Date	Sample	Democrats (D)	Republicans (R)	Spread
RCP Average	9/6 - 9/22	--	45.3	44.9	Democrats +2.8
Economist/YouGov	9/17 - 9/20	1318 RV	45	40	Democrats +5
Politico/Morn Consult	9/16 - 9/18	2005 RV	46	41	Democrats +5
NBC News	9/9 - 9/13	1000 RV	46	46	Tie
FOX News	9/9 - 9/12	1201 RV	44	41	Democrats +3
NY Times/Siena	9/6 - 9/14	1399 RV	46	44	Democrats +2
Harvard-Harris	9/7 - 9/8	1885 RV	51	49	Democrats +2

Likely voter congressional polling

Polling Data

Poll	Date	Sample	Democrats (D)	Republicans (R)	Spread
RCP Average	9/6 - 9/22	--	45.3	44.9	Republicans 3.3
ABC News/Wash Post	9/18 - 9/21	LV	46	51	Republicans +5
Emerson	9/20 - 9/21	1368 LV	45	45	Tie
Rasmussen Reports	9/18 - 9/22	2500 LV	42	44	Republicans +2
Trafalgar Group (R)	9/7 - 9/9	1081 LV	42	48	Republicans +6

Now people have long suggested that pollsters like Rasmussen and Trafalgar are biased towards the GOP. Could it have really nothing to do with "bias" and way more to do with the fact that they are pretty much always working with "likely voter" samples versus "registered voter" samples most pollsters use? Because it has been historically the case that the former are always more favorable to Republicans, while the latter is more favorable to Democrats. 

So that can become sort of a chicken vs the egg argument. Is Rasmussen and Trafalgar really more favorable to the GOP because of bias or simply because they use likely voter samples which has a built in favorability to Republicans. Is the really these two pollster, or the idea of the likely voter poll that upsets the left.

The difference between the two samples is 6.1%  as the registered is 2.8 +D while the likely is 3.3 +R. This is more than the 4% swing that the ABC/WaPo poll showed, so it would appear that there might be more than just the type of sample involved. But at the end of the day the lionshare of this difference is really how far you want to parse out the voter intentions. In the ABC/WaPo poll the difference appeared nominal (about 75% of Democrats versus 80% of Republicans were considered "motivated" to vote) but even that 5%  can make a difference of 2-3 points by itself. One must have to also look at independents and leaners to find similar intentions, and there are other factors. 

I expect more and more pollsters to switch to the likely voter model in the coming weeks and I would similarly expect some movement towards the GOP as they do.