Survey USA poll is completely unreliable, and here is why.

[adoraz]

Here are the 5 polls which have been released today:

Strategic Vision (Clinton +5)
Rasmussen (Clinton +5)
ARG* (Tie)
Quinnipiac (Clinton +6)
SurveyUSA (Clinton +18)

*released 2 days ago, but polled within the same exact time period as the others

Now, clearly there is an odd man out here. I tried to figure it out by the crosstabs, but what I found was basically Obama didn't have the same backing as in the other polls (75% of black vote? I doubt it), as well as a few other small factors such as percent of black voters and ARG. Here is why the difference is so big, its simply MoE:

Quinnipiac- 1,340 LV
Strategic Vision- 1,200 LV
Rasmussen- 695 LV
ARG*- 600 LV
SurveyUSA- 597 LV

ARG and SurveyUSA have the same amount of voters. Arg's poll is 5-6% off from the other 3 (In Obama's favor), and Survey USA's poll is 12-13% off.

Survey USA's poll sampling is less than half of both Quinnipiac's and Strategic Vision's. The top 3 polls have a 1% difference between the 3. The lower 2 have an 18% difference between the two.

All these polls were conducted the same exact time. Until someone can tell me a legitimate reason why SurveyUSA's poll is more accurate than the top 3 (which have an average of nearly 2X as many people sampled, and all just happen to be within 1%), I see no reason for people to be going crazy over SurveyUSA's numbers.

[DemocratSinceBirth]

1. The Poll Is Most Likely An Outlier But They Have A Good Track Record

ARG is a joke... I could make up numbers and be more accurate than them, ergo:

[adoraz]

5. yes I know all of this, but something to consider

Survey is ranked #1, but Quinnipiac's MoE is only 0.33 off, which is basically nothing.

Also, that is very outdated. The newer numbers may be different.

SurveyUSA is great, but I doubt they just happen to be the only ones correct about this.

[Bullet1987]

16. PPP is reliable...

[DemocratSinceBirth]

18. Doesn't Look That Way- I Know They Say They Are The Best

[mckeown1128]

2. ARG is sucky and SUSA has been the most reliable...

however... all the polls have been wrong from primary to primary.

[DemocratSinceBirth]

4. The Election Is Two Weeks Away So We Will Never Know

The "only" way to test a poll is to match it to actual results.

[adoraz]

7. I agree, ARG is terrible

but whats the excuse for the other 3 polls that all happen to be within 1% of one another? I have seen other polls from a couple days ago with Obama even ahead by a couple percent, but none with Clinton's lead being anywhere near 18%.

like you said, all have been different from primary to primary.

Almost always though, when there is just one the has such a big difference, that is the one which is wrong.

[mckeown1128]

15. Ohh don't get me wrong... I think this SUSA is whacky...

I was just pointing out that SUSA has a VERY good track record... but every poll has bad samples... and SUSA has been wrong before...


ARG on the other had has never been right... I wouldnt' be suprised if they played a bingo game to get their polling results... Christ.. even statistically they should have hitten something right by now.

[rinsd]

3. SurveyUSA has done ok but polling has been off all primary season

http://www.surveyusa.com/index.php/surveyusa-report-cards/

[my3boyz]

6. I'm not sure which one is right but

I immediately knew that the polls that had Obama leading in PA were off. I did think he was closing in a bit but I said repeatedly that he would not win the state. I believe he will lose by about 9-10 pts (like Ohio). I will be very surprised if it is by 18 pts. At any rate, I just want it over so that another big state will be gone and Clinton will finally admit that it is over! She might win PA and possibly Indiana but Obama will be our nominee.

[adoraz]

8. I agree

I think he will end up about 6-8% away from her, which would be great.

[Mass]

9. You are of course aware that, during this primary season, SUSA has been a lot more accurate than

many other pollsters.

As an Obama supporter, I prefer the other scores, but I would suggest that people do not put too much stock in these polls. Given what the polls have been a couple weeks ago, being within 10 % of Clinton would be a success for Obama and a disappointment for Clinton.

[adoraz]

10. yes, I am completely aware of that, however...

this is the first time they have been so far off from the others. there has been no other time when they were double digits off from the other polling average that I can remember.

my point is though, the virtually every poll has it between +2% for Obama and about +9% for Clinton.

If Survey is the only ones correct, I will be very shocked. I will never question them again after that.

[jackson_dem]

11. 74-24 with the black vote is plausible in PA. One word: Michael Nutter

[adoraz]

13. It seems to me that every poll from every state

has always had black votes being lower than the actual primary.

we'll see, but I am guessing he will get about 85%.

[Barack_America]

17. Lol! You don't know Philly very well, do you?

If you're insinuating that Nutter is going to be raking in the "black vote" for Hillary.

The only reason Hillary's support is so high amongst African Americans is AFSCME and Bill.

[tandem5]

12. survey usa was also an outlier in California

but they got it dead on.

http://www.realclearpolitics.com/epolls/2008/president/ca/california_democratic_primary-259.html#polls

[cbayer]

14. The only polls that are right

are those that support *your* candidate. At least at DU.