Inflated Clinton Poll Theory: Rasmussen Ups the Ante

by Chris Bowers, Mon Apr 23, 2007 at 01:15:37 PM EST

Rasmussen has upped the ante in the inflated Clinton poll theory with a poll today showing Obama and Clinton in a dead heat nationally:

Obama: 32%
Clinton: 32%
Edwards: 17%
Richardson: 3%
Biden: 1%
Dodd: 1%
Clark: 1%

In addition to the primary preference numbers, which I believe are the first scientifically random numbers that do not show Clinton alone in first place this entire election season, another aspect of the poll caught me eye (emphasis mine):
The current survey of 579 Likely Democratic Primary Voters was conducted April 16-19, 2007. The margin of sampling error is +/-4 percentage points with a 95% level of confidence.
These poll numbers from Rasmussen are in sharp contrast to recently released national polls from Fox (PDF), which showed Clinton ahead 41-20 with Gore included in the question, and ABC-WaPo, which showed Clinton ahead 41-25 without Gore included in the question. Rasmussen also stands in contrast to Fox, ABC-WaPo, and pretty much every other national survey in terms of who they include in their national primary preference poll sample.

First, let's look at who Rasmussen includes in their sample. I sent Rasmussen an email asking for more info a little while ago, and when I hear back I will post an update. For now, however, a decent amount of information is already available. A couple days ago, Mystery Pollster wrote the following after an email exchange with Scott Rasmussen:
He also confirmed that the 774 likely Democratic primary voters came from a pool of 2,000 likely voters. So last night I asked what portion of adults qualified as likelyvoters so we might do an apples-to-apples comparison of the relative "tightness" of survey screens... Via email, Scott Rasmussen tells me that while he did not have numbers for that specific survey readily available, the percentage of adults that qualify as likely general election is typically 65% to 70%...for that series. He promised to check and report back if the number for this latest survey are any different.
Rasmussen's general election likely votes are drawn from a standard, four-day sample of 2,000 participants in their daily Bush tracking poll. This is only slightly "tighter" than the potential national sample of all registered voters (73% of the adult population). It is also slightly "looser" than the potential sample of actual voters in the 2004 Presidential election (60% of the adult population). Assuming an average universe of 67.5% of the adult population for their polls, Rasmussen's latest sample of 579 Likely Democratic Primary Voters represents a universe of just under 20% of the American adult population ((0.675 * (579 / 2000) = 0.1954). Given the voter turnout numbers in both 2000 and 2004, 19.5% of the voting age population is actually a very good estimate of voter participation in Democratic primaries nationwide. While turnout in Democratic primaries in most states was actually lower than 19.5% in the past two cycles, it was higher in some very early states such as New Hampshire, and current trends indicate that 2008 will feature higher Democratic primary turnout in 2008 than in recent elections. Projecting 19.5% Voting Age Population turnout for Democratic primaries in 2008 seems quite reasonable to me.

Now, by way of contrast, here are the turnout projections in other recent national primary preference polls, and the amount by which they projected Clinton ahead. I included every April survey for which I could find data:
  • Gallup, 4/15: 50.0% of the Voting Age Population (VAP). Clinton by 8% withoutGore
  • CBS, 4/12: 39.4% VAP. Clinton by 15% in three-way poll
  • LA Times / Bloomberg 4/9: 40.6% VAP. Clinton by 10% in three-way poll
  • Time, 4/9: 38.5% VAP. Clinton by 7% in three-way poll.
  • Gallup, 4/5: 50.0% VAP. Clinton by 24% in three-way poll.
While I was unable to find results for Fox, ABC-WaPo and CNN, I have no reason to doubt that they are similarly sampling a universe of between 38-50% of the voting age population. Thus, in every case, Rasmussen shows a closer nomination campaign than other national primary preference polls, but also samples a far "tighter" universe of Democrats.

And so, one simple question emerges from all of this: is the difference between Rasmussen's national Democratic primary preference numbers, and the numbers of other polls, the result of the different universes the polls are sampling? To put it another way, does Clinton perform better in non-Rasmussen polls because those polls include a far greater percentage of "unlikely" primary voters? As I indicated earlier in the post, I emailed Rasmussen Reports in an attempt to acquire the information necessary to answer this question. Already, the one data point available to test the inflated Clinton poll theory suggests that the central thesis is correct, and that Clinton is either not ahead by the amounts most national polls suggest, or even that she is not ahead at all. This mounting circumstantial evidence from Rasmussen further suggests that the theory is valid. If the thesis is demonstrated, it would throw most existing national polling under severe doubt, and significantly alter national perceptions about the state of the campaign. Because I believe in the continuing importance of this story, I will continue to pursue it until I have arrived at a statistically satisfactory answer. I have created a new tag, "inflated Clinton poll theory" in order to create a permanent page to monitor the progress of this campaign.




Tags: Polls, Hillary Clinton, President 2008, inflated Clinton poll theory (all tags)

Display:
Permalink :: 28 Comments :: Post a Comment


Just look at Iowa and New Hampshire. (none / 0)

Clinton does significantly worse in those states than she does nationally. That's a good indicator of what's to come as voters in other states also begin to take a closer look at the candidates.


John Edwards 2008
by MeanBoneII on Mon Apr 23, 2007 at 01:26:45 PM EST

Re: Just look at Iowa and New Hampshire. (none / 0)

That would also be circumstantial, since state polls cannot be directly compared to national polls. But early indications agree with you.
by Chris Bowers on Mon Apr 23, 2007 at 01:43:22 PM EST
[ Parent ]

Karl Rove Called (none / 0)

He wants his "math" back.


by delmoi on Mon Apr 23, 2007 at 07:19:31 PM EST
[ Parent ]

Re: Inflated Clinton Poll Theory: Rasmussen Ups th (none / 0)

You want to mention I had a diary on this poll and there was a discussion going on there?


by Obama08 on Mon Apr 23, 2007 at 01:29:16 PM EST

Re: Inflated Clinton Poll Theory: Rasmussen (none / 0)

State-by-state polling has not tightened at all, if you look at, say, the aggregate of the last 10 to 15 state polls we have seen come through.  I can't believe all these other polls are wrong, virtually every single one, including a vast majority of state polls, while Rasmussen is the only one to have it 'correct' as a tie.  

Rasmussen is a GOP pollster.  He tries to stay neutral and often succeeds.  But, given the fact that Clinton is a polarizing figure especially for GOPers, it seems to play into these particular polls.   How else to explain the discrepancy between Rasmussen's poll and all the other polls?   ALL the other polls are bad, including the huge amount of state polls we have seen so far?  More likely that there is something slightly off with GOPer Rasmussen here, IMO.  


by georgep on Mon Apr 23, 2007 at 01:29:39 PM EST

Re: Inflated Clinton Poll Theory: Rasmussen (none / 0)

Yes, but those states polls could be doing exactly the same thing these national polls are doing. As I pointed out earlier, in Iowa, Clinton does 9% better among all self-identifying Democrats who are registered to vote than she does among likely Democratic caucus goers. and keep in mind that the likely caucus goers were in the registered voter sample in that poll. In other words, in Iowa, Clinton actually led Edwards by 12.5% among self-identified Democrats who were not likely to vote, but trailed by 5.5% among likely Democratic voters. That is just an enormous gap that could be screwing with a lot of these polls.

Now, I'm not saying that the theory has been proven. So far, it hasn't. I freely admit that--we need more data. Instead, I think this is a very legitimate area of inquiry into current primary polling that must be explored. Several professional pollsters agree with me about this--something strange is going on here, and we need to figure out if there is a there there.

And it is important, too. Many people base their support on the candidate or candidates who they feel are most likely to win. If inaccurate information on who is winning is presented to those people, it will heavily skew the Democratic nomination process. I intend to make sure that does not happen. I want an informed Democratic electorate.
by Chris Bowers on Mon Apr 23, 2007 at 01:42:29 PM EST
[ Parent ]

Campaign poll (none / 0)

I would think the Obama campaign would direct their people to take very credible polls to see where the numbers stands....And if their poll were showing something vastly different, i think they would have let it be known by now..Don't you think?


by JaeHood on Mon Apr 23, 2007 at 02:53:56 PM EST
[ Parent ]

Re: Campaign poll (none / 0)

Not necessarily. Some presidential campaigns don't conduct national polls during the primary season. Some simply haven't even conducted any polls yet, like Edwards. And some just don't ever release internal polls, even if they are good for their candidate. I actually have never seen a presidential campaign release an internal poll in recent years.

What I am saying is, that isn't necessarily an indicator of anything.
by Chris Bowers on Mon Apr 23, 2007 at 03:02:11 PM EST
[ Parent ]

Caucus Polls (none / 0)

What I find interesting, is that we haven't had another one of these polls...

October 12-19, 2006
Harstad Strategic Research (Environmental Defense)
Results (likely caucus-goers)

Edwards 36%
Clinton 16%
Obama 13%

Results (Democratic county chairs, vice chairs)
Edwards 40%
Vilsack 15%
Obama 11%
Clinton 8%

According to the article...

Among these Iowans most involved in party politics, the poll found even more enthusiasm for Edwards

We hear about how the caucuses can be swayed by experienced caucus-goers, but there is no repeat of this poll since October?


BlueNC - Progressive NC Politics
by Robert P on Mon Apr 23, 2007 at 03:12:11 PM EST
[ Parent ]

Re: Inflated Clinton Poll Theory: Rasmussen (none / 0)

As far as the idea that Rasmussen would downplay Clinton because she's so hated by Republicans, I happen to think the opposite theory makes more sense: that they would overplay her strength as the inevitable nominee since a Clinton candidacy is about the best thing the Republicans can hope for in next year's race.

Take a look at Fox's polls so far; why else do you think they insist on including Gore, even now, than that it helps Clinton's lead in such polls?


by the wanderer on Mon Apr 23, 2007 at 01:44:38 PM EST
[ Parent ]

Re: Inflated Clinton Poll Theory: Rasmussen (none / 0)

Excuse me if I'm repeating, but you have only 22 - two - competing methods here. The fact that a lot of polls using the same sampling method (and likely voter screen) show roughly the same results during the same time period, proves nothing except that they use the same method.

Rasmussen is using a different method, as Chris has explained so well. The matter of the dispute is which method is better, not which method is more often use.

Why isn't Rasmussen's method used more often? Hint: It is much more expensive.


by PoliticGeek Pro on Mon Apr 23, 2007 at 02:31:39 PM EST
[ Parent ]

Re: Inflated Clinton Poll Theory: Rasmussen (none / 0)

ZOGBY'S NUMBERS ALSO HAVE IT MUCH CLOSER AS WELL AS THE NEW TIME AND GALLUP POLL.

Strategic Vision did a poll recently in GA which showed Clinton ahead of Obama by just 3 pts 25-22.

Rasmussen had the best record in 2004 and 2006 elections of all of the pollsters. Their final totals for Bush Kerry were within a couple of tenths for the final outcome.

Yes they are republican leaning but have been much more neutral since 2000 when they had the worst record in polling. For business and credibility they need to be accurate.


by BDM on Mon Apr 23, 2007 at 02:51:27 PM EST
[ Parent ]

Using your theory (none / 0)

Doesn't that mean that John Edwards number (behind Clinton and Obama) are also then very underated, looks like that would apply to his lower numbers also.

Making him underestimated.


Check out the New Progressive Blog EENRBLOG
by dk2 on Mon Apr 23, 2007 at 01:45:45 PM EST

Re: Using your theory (none / 0)

Yes, I think that Obama, not Edwards, will be the primary beneficiary if my theory is proven to be correct. I do not think that Edwards is experiencing any noticeable national deflation in his numbers as a result of these polls, but I do think Obama is.
by Chris Bowers on Mon Apr 23, 2007 at 02:12:59 PM EST
[ Parent ]

Chris, how do they (none / 0)

determine likely voters? It can't be as simple as asking them if they plan to vote, can it?


by david mizner on Mon Apr 23, 2007 at 01:51:29 PM EST

In case Chris doesn't get to it (none / 0)

I think it has something to do with whether or not they've voted in recent elections, the theory being that if they voted previously, then they'd be more likely to vote in the next big election. And this is the question that is asked them. Just something I read once.


by mihan on Mon Apr 23, 2007 at 01:59:43 PM EST
[ Parent ]

Well then (none / 0)

it could undercount someone like Obama, who's bringing new people in.


by david mizner on Mon Apr 23, 2007 at 02:03:03 PM EST
[ Parent ]

Re: Well then (none / 0)

That's what I thought about Kerry in '04. It could very well be true, for Obama or anyone bringing in new voters, but those numbers would theoretically be reflected in any sample of simple 'registered' voters versus 'likely'.

I'm sure the Clinton camp isn't particularly happy about these numbers in particular, though.


by mihan on Mon Apr 23, 2007 at 02:10:13 PM EST
[ Parent ]

Re: Chris, how do they (none / 0)

I'm trying to find that out. More info when it comes.
by Chris Bowers on Mon Apr 23, 2007 at 02:11:36 PM EST
[ Parent ]

Likely voter screens (3.00 / 2)

The Mystery Pollster had some details on likely voter screens for the 2004 general election, which would be too loose for the primary, but you can get an idea of the methodology from the examples. Sometimes, it is as simple as asking whether the respondent plans to vote.

I would also point out a second data point for comparing primary voters and more general voters. I conducted three identical polls in Texas, with the one exception of different populations. Based on the verified voting history in the Vote File, one population was general election voters and the other two were regular primary voters. For the general election voters, I asked if they would vote in the primary and then which primary. The other two populations were voters with a history of voting in one of the two primaries. I still asked the 'which primary' questions of the primary voters, so the polls would be conducted exactly the same way.

Among the general election voters who claimed they would vote in the Dem primary, Clinton scored eight points higher than among the past primary voters. On the GOP side, McCain also did better among the general election voters than among the primary voters


Texas Economics
by IVR Polls on Mon Apr 23, 2007 at 02:27:06 PM EST
[ Parent ]

Re: Inflated Clinton Poll Theory: Rasmussen Ups th (3.00 / 1)

Did anyone ever do any analysis of whether Howard Dean polls were inflated in 2004?


BlueSunbelt.Com Netroots for the Sunbelt states robwire.com My personal blog
by robliberal on Mon Apr 23, 2007 at 02:13:53 PM EST

Re: Inflated Clinton Poll Theory: Rasmussen Ups th (none / 0)

I didn't know nearly as much about polling then as I know now. In retrospect, I certainly think that they were inflated. However, I just did not know all of the things that I know today, and so the need to make such an investigation, or even how to conduct an investigation, was beyond me.

I don't remember anyone else doing so either.
by Chris Bowers on Mon Apr 23, 2007 at 02:20:13 PM EST
[ Parent ]

I crunched CA State Poll Numbers (none / 0)

Numbers could only be compared at the Registered Voter level, but tell me it doesn't look pretty similiar


by Bob Brigham on Mon Apr 23, 2007 at 02:49:29 PM EST

Dean's numbers simply collapsed (none / 0)

whether his polling nubers were inflated is irrelevant he plunged in Iowa like a stone in the last weeks and nationally fell of the face of the earth afterwords,


Obama! because 51% isn't enough!
by nevadadem on Mon Apr 23, 2007 at 02:52:20 PM EST

This is complicated... (none / 0)

by the fact that there is no national primary. Perceptions of the race can change very quickly as the results from early states come in. One need look no further back than 2004 to understand that national polls polls mean little.


by rebop on Mon Apr 23, 2007 at 04:59:17 PM EST

Re: This is complicated... (none / 0)

Ah, but they do matter in terms of the national narrative about the race, and thus the resulting scramble for activists, money, media, credibility and other forms of support.
by Chris Bowers on Mon Apr 23, 2007 at 05:10:02 PM EST
[ Parent ]

Wishfull thinking (none / 0)

Kos thought Gray Davis would win re-election.

Trying to discount Clinton's poll numbers reeks of Karl Rove's "math".  She's ahead now. The important thing is to bring down those numbers as much as possible between now and next january.


by delmoi on Mon Apr 23, 2007 at 07:18:28 PM EST

Re: Wishfull thinking (none / 0)

I'm glad you have offered an insightful, statistically based contribution to this discussion. I guess because Kos was wrong about the California recall election four years ago, it somehow means that my analysis has no validity. Truly biting.

Maybe you think Mystery Pollster is engaging in "wishful thinking" too.
by Chris Bowers on Mon Apr 23, 2007 at 07:34:50 PM EST
[ Parent ]


You are not logged in.

In order to post a comment, you must be logged in. If you have a member account, please log in to comment.

If not, you can make an account right here. It's quick and free.

Login
Username:

Password: