Pollster.com

wow - what differences. i have to go to a meeting so i cannot look at the samples. i'm sure someone will analyze them before i return. :)

What is especially puzzling is that the polls are wildly divergent on the Democratic side, but reasonably close on the Republican side. I don't know what that means, but it might help point the way to an explanation of why they differ, assuming they use similar methodologies for both parties.

What is especially puzzling to me, as a lay obserer, is that these polls are wildly divergent on the Democratic side, but reasonably close on the Republican side.

I don't know what that means, but it might help point the way to an explanation of why they differ, assuming they use similar methodologies for both parties.

Rasmussen and ARG sample were around 600 - TCUL- 400.

The ARG poll is the first I've seen Obama ahead in TX so I think it's the outlier- though the race has tightened some probably.

ARG has African Americans as 22% of likely voters, while they comprise only 11% of the state's population. It would seem the poll is heavily weighted.

Moishele,

First off, are they 11% of the population or 11% of the democratic population? That would be a BIG difference.

Second, the strange Texas primary system will end up giving more delegates to African-Americans than proportional and less to latinos. Maybe its weighted for that too?

Moishele:

Remember that this is the Democratic primary and most of the African-Americans probably are Democrats.

Still, it is hard to say anything here except that Obama is clearly doing better than he was a few weeks ago.

Considering that African Americans are about 90% Democrats and usually turn out at a slightly higher rate than whites and a much higher rate than Hispanics in Texas and seem likely to turn out an even higher rate than normal this time, 22% doesn't seem off to me.

Moishele -- though African-Americans may be only 11% of the overall Texas population, might they not be a much larger percentage of the Democratic party in Texas?

Indeed, my assumptions (untroubled by facts) would be that Democrats probably make up less than 50% (probably less than 40%) of Texans. But African-Americans are probably almost all Democrats. Thus, a group that is 11% of the general population might represent a much larger percentage of the Democratic population -- the 22% figure seems right in that context, if not a bit low.

One thing that does seem questionable about the ARG poll is that it assumes 47% will be male. I believe that is much higher than most primaries this year and would obviously favor Obama.

22% seems a little bit high but not overwhelmingly high. I would guess more like 18-20%. African Americans are 90% democrat so in general, you can multiply their percentage of the state population by 1.6 to 1.8 to get their percentage of the democratic vote. However, even with this being the case, I find it hard to believe that Obama is ahead in TX, as much as I would like to believe that was the case.

Moishele - Blacks are typically members of the Dem party - so the fact that their numbers are disproportional to their overall representation in the state is not unusual. In some of the southern states, for example, blacks were almost 50 percent of the voters - although they clearly are not 50 percent of the population in those states.

Volatility in the democratic side seems to have been a pattern thus far. I think Mark/Charles have commented on this previously, noting several factors. The one that makes the most sense to me is that both Clinton and Obama are rather highly esteemed among Dems. This likely leads to more "fluid" preferences among voters; not only are the undecided voters more easily swayed, but even the voters who express a preference probably are not as solidly rooted in their choice as you might expect.

There is an unfortunate tendency among laypeople to look at highly variable poll results and immediately ask "What or who is wrong?". The reality is often that nothing is wrong, but that the polls are actually _accurately_ reflecting a large amount of variability (i.e. indecision) among voters.

My recollections of the CA polling leading up to Super Tuesday seems to reflect this. There were a large number of pretty different looking polls, which suggested to me at the time that things were strongly in flux and that the race was fairly close, in the sense that large numbers of relatively uncertain voters could swing the election either way. In CA, they broke comfortably (but not overwhelmingly) for Clinton.

Along these lines, I wonder if Mark or Charles would like to weigh in on two issues in more detail:

1.) What statistical evidence can we muster to show that the Dem primary race this season has been more "volatile" than usuall? If it has been more volatile, how exactly, and what effect has this had on polling and the race in general?

2.) Granting that the Dem race has become fairly volatile post-Super Tuesday, how useful _really_ are the state-specific polls we've seen since then? Specifically, given the spacing of subsequent contests, Obama's string of victories and the fact that a poll's field dates will likely lag behind or straddle recent important events, isn't is possible that we won't really have a clear (state specific) polling picture until (my guess) 5-7 days after HI/WI

I caveat with the state specific note because I think we've seen pretty convincing evidence that Obama's recent successed have translated into an upswing in his national numbers at least...

NOw why would people with computers on a forum about data and surveys guess????? Anyways her's the data.

Texas 2004 Primary Exit Poll Data
Whites: 52
Blacks: 21
Latino: 24
Asians/Other: 2

Census Data from Texas
White: 71
Blacks: 11
Latino: 35
Asians/Other: 2

That being said, Latino's in California came out in double the numbers ever seen historically. If that's any indication of turnout, or that fact that in a competative race whites may also be more likely to turn out and that in 2004 the Republican nomination was uncontested mean these turnout %'s by race will probably be inaccurate.

Blacks also have not turned out in greater numbers than their percentage of the population or in previous contests this primary cycle. Yes thier vote has overwhelmingly gone to Obama, but it did not increasein percentage of the electorate. In fact, it went down.

Also, one thing to keep in mind about the ARG poll is that they sampled INDEPENDENTS, which overwhelmingly went for Obama. Texas is an open primary. Rasmussen only samples registered democrats. In this case, I think Rasmussen is the outlier. I'm guessing the race is pretty much tied at this point.

Listen, you Obama Bots, Hispanics make up about 25 to 30% of Texas. Most Hispanics support her over Obama. ARG is a known republican organization, They had Hillary 13 pts ahead in Iowa the day before the race and look how that turn

Rasmussen is a republican pollster and even he states that she is Leading in Texas so stop with the poppycock explanations please. You make yourselves sound dumb.

A lot of people say that the polls in calfornia was not correct as many of them show Obama ahead just before the election date. Did any get a chance to analyze that it has nothing to do with absentee ballots cast much much earlier? I heard almost >20% of the ballots were absentee mailed long before February. Is there a way to put that aside and see if the polls were correct for election day voting?

Saw this analysis of the crosstabs released by the Texas Credit Union League...it also refers back to the 2004 exit poll data for comaprison.

http://ccpsblog.blogspot.com/2008/02/analyzing-texas-polling-on-march-4th.html

John Barnes,

What exactly are you implying by saying that ARG and Rasmussen are "Republican" pollsters? Why does it matter? If it does, why do they have COMPLETELY different polls? If it doesn't, why bring it up at all? What does ARG's failure in Iowa have to do with anything?

S.b.--

Could you provide a link to support the last point? I'm having a bit of trouble finding the data. Also, I wonder if the huge Latino turnout and lopsided preference for Clinton in CA was also reflected in other states with significant Latino populations.

Finally, it has already been noted that the apportionment seems to favor AA over Latino voters...it should also be noted that some delegates will be chosen by caucus; Obama's success in caucuses has, shall we say, not gone unnoticed.

Mike,

Can I get a source for this info:

"Rasmussen only samples registered democrats. In this case, I think Rasmussen is the outlier. I'm guessing the race is pretty much tied at this point."

Thanks

@Akilis:
In California, SurveyUSA included mail-in as well as election day voters, and they were on the money.

The ARG poll suggests Senators Clinton and Obama are tied among Hispanics - 44-42! That could be partly why they are seeing an Obama lead.

OK, this article contains some relevant info:
http://www.cbsnews.com/stories/2008/02/05/politics/main3795497.shtml
Clinton's edge among Latino voters was huge in CA but smaller in AZ (55-41).

The article claims that black turnout was heavy across the nation, but offers comparative numbers only for DE, where AA voters went from 16% of the primary electorate in 2004 to 28% this year.

I'm not sure what is more troubling for Hillary - the rate at which the polls are tightening up, or that Obama hasn't even campaigned there yet?

Wow, this certainly doesn't bode well for her at all. And in light of two powerful Unions endorsing Obama, March 4th may very well be the end of the line for her if these current trends continue.

I would like to see a bit more information on the demographics of the poll as well as when voters' decisions were made.

This is really quite historic the way Obama's campaign has "upset" the machine that was Hillary Clinton. I guess it goes to prove what an amazing country we live in when people get together to accomplish something.

Bob Evans - You must be really worried. Once she takes WI. March 4th is the day Obama bites the dust and the media will dump on him. Wake up folks!

Mike, are you sure Rasmussen sampled Dems only? Their poll says "577 likely Democratic Primary voters," which grammatically could well include indies who've indicated an intention to request a Dem ballot.

http://www.burntorangereport.com/showDiary.do?diaryId=4972 - REALLY interesting analysis of the Texas Credit Union League poll. Summary: Obama wins 8 delegates more than Clinton even though he's down 8% in the poll. This doesn't even count the caucus portion of the Texan contest. I want to reiterate this: this is an analysis of the most reasonable poll that came out today NOT of the poll that gave Obama a lead in Texas.

Remember: this is before he starts campaigning there, before his union endorsements kick in, and before next week's (probably) victories kick in.

If Clinton wants Texas as a firewall.. she better start praying for a miracle.

There's something else about the ARG poll that I don't know if anybody has pointed out yet, so I'll do it now.

ARG has now done two polls of the Texas Democratic race, the Feb. 2008 poll and the March 2007 poll. The March 2007 poll had Clinton leading Obama by only 2 percentage points, 34-32. Other polls of Texas at the time gave Obama a much smaller number, like in the teens, presumably because Obama was still considered unknown at the time.

That leads me to believe that ARG may be doing something that is inflating Obama's numbers in Texas.

Shane,

As someone pointed out, I misread the Rasmussen poll which said "Likely Democratic Primary voters" which very well could include Independents. Because crosstabs are unavailable expect by subscription, we'll never know.

I will, however, make this point.
In the TCUL poll, the crosstabs indicate that they polled "Strong Dem" voters and "Soft Dem" voters (www.tcul.coop). "Soft" may include Dem-leaning independents ? ARG explicitly included Indies. Despite the 17-point spread between these two polls, they seem remarkably closer, and both seem to indicate substantial inclusion of Independents.

That said, the best poll is probably the TCUL poll, but a more detailed crosstab of what the exact percentage of Independents was would be helpful.

ARG indicates that Indies go almost 3:1 for Obama. If Rasmussen did include Indies, and they went similarly in his poll, that would indicate even lower support from core Dem voters, which seems improbably considering the crosstabs from TCUL.

Does this logic make any sense?

By the way, I'm not an Obamabot just trying to hype him up. I'm objectively interpreting divergent poll results.

RS:

You are right SurveyUSA was spot on California and Clinton had a 17% lead in mail in voting but only 6% lead on voting day. I wish they could break it down to latino votes too. That way we would know the real composition of latino vote on the voting day. Does anyone know how many people already voted in Texas, Ohio and PA? The later people vote the more chance Obama has to make his positions known.

Roy just look at the exit poll data for the most recent contests in the Potomac. The black percentage of the vote went down slightly, very slightly. From CNN and CBS web sites. In Viriginia, 33% of the primary voters were black in 2004. In 2008 it was 30%. In Maryland 35% of voters were black in 2004, which went up to 37% in 2008. These are survey's of course. Also the blcak population is increasing in these states by about 1% in this time. So there is no evidence to assume that the % of black voters has gone up in this election cycle. None that I have seen. The black voter population has remained within the margin or error, population increases, or gone down slightly.

I'm sure you can check all the states on CBS and CNN, maybe even ABC. The difference this year I believe will be an increase in white and hispanic voters in Texas, because the race is still in play and because even Republicans have a choice. Huck an MAc are tied right now. Huck may drop out, which would change things, but one assumes some blacks are Republicans in Texas and their % of the vote will actually go down this cycle for the above reasons.

I just went through the TCUL crosstabs again, and they do include Independents, who split 59-26 for Obama. I guess the major difference with ARG is the percentage of the total electorate they will make up.

TCUL has core Dems 53-37 for Clinton, which is REMARKABLY SIMILAR to Rasmussen's 54-38.

Just a thought...

There are no registered democrats or republicans in Texas. Everyone arrives at the polling location technically independent. At that point you have to choose. You can only vote once.

In addition, 25% of the delegates are awarded in caucuses after the polls close.

Early voting in Texas begins 2/19...

There's also without a doubt some bounce in some of these polls. When checking the dates, The polls that include the 12th and particularly the 13th are the highest for Obama. BOunces go up then they come down. That's why they are called bounce. Rasmussen, had Obama up by 12% yesterday, only 8% today. The bounce is coming down and their poll of Texas was done yesterday. Polls including Tuesday and Wednesday, maybe even Thursday will be skewed by this. Wait a few days.

Clinton will certainly do better in Wisconsin than the other primaries. She will probably also do well in Hawaii as the democrats there are old, white, Asian, military and behind her. YEah she didn't do well in Maine, but i think she will in Hawaii.

Anyways the bounce and the hype will go away, then we'll see.

Akilis,

Not sure about Ohio, but early voting doesn't begin until February 19th here in Texas.

Also, I would not underestimate the anti-Clinton feeling in Texas among many true independents (though technically everyone's an independent in Texas).

I appreciate the Texas Credit Union League poll (even though it kind of reminds me of those California anti-gaming initiative polls that wrongfully showed Obama surging on Clinton in that state) because it helpfully breaks down Clinton and Obama's support into the different regions of Texas.

According to the Credit Union League poll, Clinton leads Obama in the West, South, and East parts of Texas; Obama leads Clinton in the Central and Houston areas of Texas; and Clinton and Obama are tied in the Dallas/Ft. Worth area of Texas.

These regions do now exactly match of with the Texas Senate districts, but they're still helpful.

Good news for Clinton supporters: Many of the Texas Senate distrcis in the Houston area are represented by conservative Republican Senators, so it's possible that Obama's lead in the Credit Union League poll's "Houston area" is confined to one or two Senate districts where Democrats are actually influential.

"I appreciate the Texas Credit Union League poll (even though it kind of reminds me of those California anti-gaming initiative polls that wrongfully showed Obama surging on Clinton in that state) because it helpfully breaks down Clinton and Obama's support into the different regions of Texas."

I don't see how this poll shows Obama surging. Its about the same as the Insider Advantage poll, its right in the middle of the two extreme polls (Rasmussen and ARG) and it seems consistent with the IVR poll from a few weeks back. If I was a betting man, I'd be on it being just about right (maybe a point or two more for Clinton).

Folks:
Much of the difference between TCUL and ARG is the Hispanic split. I posted the ARG numbers earlier - 44-42 Clinton. TCUL says 63-32 Clinton, with Hispanics 32% (higher than ARG?) of the voting public and African-Americans 17% (lower than ARG?)
I used the TCUL raw data and instead of 63-32 Clinton, split Hispanics with ARG's 44-42. I got an Obama lead of 45-44 overall.
This increases if the African-American vote-share is increased and Hispanic vote-share is decreased (d-uh!)

I am a little worried about the use of 400 people to represent Texas... But the race is definitely tight, apparently.

By the way, TCUL has an early-voter/election day voter split - early-voters split almost evenly for Senators Obama and Clinton.

"the bounce and hype will go away" - I wonder how much of this is wishful thinking ;-)

When did Texas stop party registration??? I lived there until 5 yrs ago and was a registered Democrat.

There could be some utility in analyzing nearby AZ and NM Latino voter preferences in the current primary season.

Clinton won Latinos 55-41 in AZ (where the governor endorsed Obama), 62-36 in NM. I would be very surprised if they split the TX Latino vote. But, I am surprised Bush "won" two elections...

*data are from CNN exit polls.

RS: Perhaps I misunderstood your post but are you saying that the ARG Hispanic number is more likely correct than the TCUL one? If so what are you basing your assumption on?

I'm wondering about one of the orange dots (for Obama) in the chart you see when you click on "TEX DEM" on this site's main page. It looks like it is above 45%, but I see no poll in Texas showing Obama above 45%. Let me know what you guys think.

Kaitlyn

never mind, see it now.

@ richard pollara:
No, I am not saying that ARG has the correct numbers; I think it is more likely they are wrong (much as I might hope otherwise!) Just that ARG shows an Obama lead, unlike other polls, because of ARG's demographic split.

@Kaitlyn:
ARG has Obama up 48-41.

I wonder if there is an analytical chart that can quantify Obama's momentum as a function of days campaigned. That, indeed, would be facinating. As well as an analysis of whether Clinton campaigned there as well and for what duration.

Although Obama is campaigning in Wisconsin right now, he has received a nearly unimaginable "bump" in polls in Texas - where Hillary is campaigning! Wow. This is something quite unexpected in terms of voter mindset and preferences I must say.

What would also be interesting - if any of these pollsters would mind asking - is what sort of dynamic is going on here? Are voters switching from Hillary to Obama, or is Obama bringing in so many NEW voters, that Hillary's support has not waned in size but in percentage due to the onrush of supporters for Obama. That would be quite interesting as well.

I left Texas for 8 years, but there was no party registration in the 80's and there is none now. If they did between '96 and '04 I can't say...but it's not even an option on the registration form.

http://www.sos.state.tx.us/elections/forms/vr17.pdf

Bob Evans, there is nothing surprising about catching up Hillary and his electoral machine, when you are back by the mainstream, left, and right wing media.

Let's not pretend that Obama is an underdog. An underdog would be John Edwards, whom the press hates with passion and whose haircuts and other irrelevant issues are always blown out of proportion.

Let alone Hillary, who is hated by the media.

I am not an Obamabot, but I think Clinton's chances in Texas are overstated. Hispanic voters in Texas are much more Conservative than elsewhere, following considerable outreach by, among others, George W. Bush.

In 2004 they voted 50-50 Bush-Kerry, so many Hispanic voters might not be inclined to vote in a Democratic primary.

Hey folks, the AA turnout didn't go up greatly in percentage, but did in actual numbers. Turnout from all groups has been dramatically up on the Democratic side. The percentage changed little but the actual numbers of AA voters showing up were much higher than in the past

The CNN 2004 exit polls are wrong when it comes to ethnicity. Voter file analysis of that and most TX Dem primaries breaks down to ~1/2 white, 1/3 latino and 1/6 african-american. TCUL is pretty close to that.