As the media begins to hype polls for the November elections, one thing stands out to anyone that compares the polls. There is a huge variance between the different polling outfits.
For example, Real Clear Politics, a group that averages the polls and tries to use that average as a measure of how the public may vote in November shows a wide margin between polls in the President’s approval rating as illustrated below.
President Obama Job Approval
|RCP Average||8/5 – 8/24||–||45.4||50.3||-4.9|
|Gallup||8/22 – 8/24||1547 A||42||51||-9|
|Rasmussen Reports||8/22 – 8/24||1500 LV||44||56||-12|
|Reuters/Ipsos||8/19 – 8/22||1063 A||45||52||-7|
|Time||8/16 – 8/17||1002 A||46||45||+1|
|Associated Press/GfK||8/11 – 8/16||1007 A||49||50||-1|
|FOX News||8/10 – 8/11||900 RV||43||49||-6|
|CNN/Opinion Research||8/6 – 8/10||1009 A||47||51||-4|
|NBC News/Wall St. Jrnl||8/5 – 8/9||1000 A||47||48||-1|
As you can see, the results go from a low of -12 in a poll by Rasmussen to a high of +1 in a Time poll. A 13 point swing.
Why the huge difference?
One of the main factors is the likelihood that the respondents will actually vote in the election. The column showing the number of voters polled also shows one of three identifying characteristics of those polled.
“A” denotes the respondents were Adults, “RV” means Registered Voters and “LV” is used for Likely Voters.
The determination of who is a “likely voter” is revealed by their answers to certain questions. Some come out and ask if you plan to vote in November, others are less direct, asking if you voted in previous elections, primaries, etc.
And the accuracy varies a good bit between a sample of Adults and those likely to vote. The assumption is that the likely voters pay more attention to the news, are better informed and actually motivated to cast a ballot.
So, if Likely Voters are a more accurate way to judge an election’s outcome, why doesn’t everyone use only Likely Voters?
Well, some polls are not looking for accuracy, they simply want to sway public opinion, or at least public perception. Rasmussen is concerned with accuracy. This is their bread and butter. Most of their polls use only likely voters, and they are very strict on determining who is likely to vote.
Others, like Time, a left wing publication, are likely more interested in their left wing agenda than an accurate prediction.
Fordham University analyzed the accuracy of 23 polls in the 2008 presidential election. Rasmussen and Pew were #1 and #2 respectively. Looking at the bottom of the list, it seems the more liberal outfits – Newsweek, ABC, CBS, NBC – all overestimated the size of Obama’s victory.
Now most of these pollsters shifted to Likely Voters towards the election, which should tell you something to begin with, as far as accuracy. And they will shift in this election as well. Some of them.
More people begin to pay attention after Labor Day, and more polls shift to the more time consuming method of polling only likely voters as well.
But based on the 2008 results, it is pretty clear that Rasmussen is the least politicized of the pollsters and is more interested in accuracy than influence.
So, most polls using Likely Voters have more credibility, but Rasmussen will again prove to be the most accurate.