Scott Rasmussen has replied to Andrew Sullivan’s post about poll discrepancies which I blogged about earlier today:
I am fascinated with those who see a big difference between our Job Approval polling and other polls. In fact, there is no significant difference.
For a Democratic President, polls of all adults are typically going to show results about 3-5 points higher than polls of Likely Voters. That’s because some segments of the population that lean heavily Democratic are typically less likely to vote than others (i.e.–adults under 30). If you look back at a comparison of Rasmussen Reports daily tracking of Likely Voters with Gallup’s daily tracking of Adults, you see the same broad trends and most of the time a gap in that 3-5 point range.
Due to statistical noise, there are times when the gap is bigger and other times when it is non-existent. There have even been times when our Likely Voter sample showed a point-or-two higher than Gallup. But, those are the exceptions.
It’s also worth noting that our polls tend to show less volatility than other polls. During Election 2008, our polls consistently showed Barack Obama with essentially the same lead every single day for the final six weeks of the campaign.
By the way, the same logic applies to other polls. A week ago, a Washington Post survey of adults found that 45% favor the health care bill working its way through Congress. At that time, we found 42% support (our new weekly update shows 45%, but the range has been steady between 41% and 46% since July).
That should be that, I guess.
Except, of course, that Andrew still doesn’t get it.
[...] about Andrew Sullivan as of October 26, 2009 Scott Rasmussen Replies To Andrew Sullivan – belowthebeltway.com 10/26/2009 Scott Rasmussen has replied to Andrew Sullivan’s post [...]
As someone else pointed to, and I alluded to earlier:
has nothing to do with their choice of likely voters. The problem stems from them consistently rating Republican Party ID (among all adults) higher than it actually is. Look at this table (below the chart): Over the past four months all pollsters except Rassmussen have shown Republican Party ID in the low to mid twenties, yet Rassmussen always has them in the low thirties.
None of these numbers include ‘leaners’ so you can’t explain it away by assuming that more independents lean Republican, or that other pollsters are calculating leaners the wrong way.
Rassmussen consistently polls Republicans at 5-10% higher party ID than other pollsters, and as a result, they weight their polls more towards Republicans respondants, biasing them against Democrats (since registered Republicans are obviously the most likely to be knee jerk anti-Democrat)
Edit: There is a possibility that Rassmussen only uses ‘likely voter’ party ID to weight, but as Rassmussen himself says above, he finds that ‘likely voters’ give about a 4 pt net bump to Republicans. Pollster’s ‘likely voter’ margin is slightly larger than Rassmussen’s ‘all adults’ margin, resulting in about a 5% bias for Republican opinions in his polls. The bias gets worse if he uses a mix of the ‘likely’ and ‘all adult’ numbers.
Not sure why he’s doing so, but it is consistent. personally, I don’t think it’s an accurate reflection.