« Poll: Obama Ahead By Seven Points In North Carolina | Home | Zogby: Obama Ahead By 16 In North Carolina, Tied In Indiana »
SurveyUSA: Obama Ahead By Six In Oregon Primary
A new SurveyUSA poll of Oregon shows Barack Obama ahead in this upcoming primary, but the result is within the margin of error. The numbers, compared to the previous poll from April 7:
Obama 50% (-2)
Clinton 44% (+4)Sample size: 650 likely primary voters.
Margin of error: ±3.9%
The Oregon primary will be held on May 20, the same day as the Kentucky primary, which Obama looks likely to lose by a wide margin.
Advertisement
Comments (54)
Within margin of error? No, it's not.
May 2, 2008 12:31 AM | Reply | Permalink
Does anybody have a source on this margin or error thing? I hear it both ways.
May 2, 2008 7:11 AM | Reply | Permalink
http://en.wikipedia.org/wiki/Margin_of_error
My read of the article is that the confidence level of the poll is for anywhere within the margin of error. The confidence level isn't described here, but let's assume it is 95%. Then, with 95% certainty clinton would get anywhere from 40 - 48% of the vote, and obama would get anywhere from 46 - 54% of the vote.
Note, that is a statistical measurement, so it is also possible that does not reflect how people will actually vote.
May 2, 2008 11:22 AM | Reply | Permalink
You have to be ahead by twice the margin of error in order to have a lead outside of it. The number is the margin of error for each candidate's individual number.
May 2, 2008 12:34 AM | Reply | Permalink
I stand corrected.
BTW:
New Poll-Obama up 16 in North Carolina/tied at 42 in Indiana
Obama-50-Clinton-34% in NC
Obama-42%-Clinton 42% in IN
http://www.zogby.com
May 2, 2008 12:41 AM | Reply | Permalink
You might want to look into that margin thing, seveal peoople I just spoke with say it's not twice the margin as you stated.
May 2, 2008 12:59 AM | Reply | Permalink
Eric,
Although it's a fairly common mistake, you've misrepresented the idea of "margin of error", essentially doubling it.
May 2, 2008 1:10 AM | Reply | Permalink
Meanwhile, you know those ten big delegates that Hillary got in Indiana, in one of her biggest victories in months?!
And how they cut into Obama's big lead, albeit in not enough numbers for the Clinton campaign to even get close to catching even?
Well, they're gone. Just checked the latest delegate count... and it says that Obama has made all of them up in superdelegates between Pennsylvania and now.
It's so fitting after all of her campaign's negativity watching Hillary get sold down the river, because her party's leaders don't want her anywhere near a position of power.
Hillary's tears taste like the sweetest nectar!
May 2, 2008 12:35 AM | Reply | Permalink
Disabled activists (most of them in wheelchairs) arrested outside McCain's Capitol Hill office.
http://www.cbsnews.com/stories/2008/05/01/politics/main4060988.shtml
Arresting people in wheelchairs for exercising their right to assembly?
John McCain can suck shit from an asshole.
May 2, 2008 12:39 AM | Reply | Permalink
Zogby has new polls out that aren't as optimistic for Her Highness, but I'm sure Eric wouldn't stay up late to post those.
May 2, 2008 12:42 AM | Reply | Permalink
He shouldn't really bother. Zogby is a joke.
May 2, 2008 12:45 AM | Reply | Permalink
They were dead on in PA.
May 2, 2008 12:46 AM | Reply | Permalink
Oh God, you're right. They were, weren't they? I remember.
After being a joke the entire primary, Zogby was the one to get the margin right in PA! I love it!
May 2, 2008 12:48 AM | Reply | Permalink
There's a good reason for Zogby being off so badly early on... their methodology was probably more complex.
"Complex?! But shouldn't that be better?!"
Under normal situations, yes. But when you have exceptional turnout trends, and the turnout trends vary so much between states, no.
Case in point: There was a huge black turnout in South Carolina... way above average. The most complex polls looked at historical data and failed to see the full extent of the turnout... whereas those polls which were more simple and didn't try adjusting their polling results as much -- such as SurveyUSA -- were more accurate, because they weren't effectively underestimating the turnout.
But after South Carolina, the polling companies with complex polls made the opposite mistake in states like California. They overestimated the percentage of black turnout, while assuming that Latino turnout would be historical. Wrong again on both count.
Think of it a bit like artillery... the big polling companies are more complex, but if they don't know which way the wind is blowing and how hard, it's difficult for them to hit the target. So they'll overshoot. And undershoot. Until they've finally started to figure things out, at which point, they'll be the most likely to be accurate.
May 2, 2008 1:04 AM | Reply | Permalink
Gosh, SUSA was actually worse than Zogby in SC (although both stunk pretty badly). SUSA's last SC poll (released two days before voting) predicted Obama by 13. Zogby's last poll (released the day before voting) predicted Obama by 15. In reality, it was Obama by 29. In other words, Zogby was 2 pts closer to the correct number, although both were off by double digits. Nobody did very well in SC.
May 2, 2008 11:34 AM | Reply | Permalink
IA's a joke but didn't stop em from pimping that one. Oh, did I say pimping?
May 2, 2008 1:14 AM | Reply | Permalink
Oregon being in the Northwest these numbers appear to contain a bit of wishful thinking. Now, Kentucky is another story, absolutely. Although, they likely will counter balance each other, they will perform as they have for the past elections cycles.
May 2, 2008 12:49 AM | Reply | Permalink
...but hey, that doesn't stop us, does it?!
May 2, 2008 1:22 AM | Reply | Permalink
Given the significant difference between the Zogby poll and the last, surprisingly low result for Obama in that other recent poll, I can only assume that Zogby feels that the black turnout is being underrepresented in the previous models... which seems like a pretty reasonable assumption, if NC is anything like SC.
This time around, I suspect they've done some things with their polls to try and make sure that this is the case.
May 2, 2008 1:26 AM | Reply | Permalink
I like Obama, but Zogby shouldn't even bother polling these primaries. His past performance is worse than throwing darts at aboard. He correlates negatively with the actual results in most of his recent polling. This poll result is almost a guarantee that it will be very, very close.
May 2, 2008 10:07 AM | Reply | Permalink
Actually, you're wrong, Eric.
The margin of error is calculated as one figure, not two, and indicates a 95% likelihood that the result will be, in this case, within the ±3.9% confidence interval as of when the poll was taken.
A result that is greater or less than that amount is considered to be "outside the margin of error".
May 2, 2008 12:46 AM | Reply | Permalink
Waiting for Zogby to be posted...
)))))))))))))))-crickets-((((((((((((((((
May 2, 2008 1:32 AM | Reply | Permalink
Actually Eric, Hillary only gained two points from the last one, not four. It was 52-42
May 2, 2008 1:53 AM | Reply | Permalink
We here in Oregon will deliver our state for Obama with margins to mollify the bloodbath in Kentucky! Puerto Rico has nothing on us! Except more delegates.
Is this over yet? Please be over. For the love of Santa Claus, please kill me.
May 2, 2008 2:34 AM | Reply | Permalink
I think Eric is confusing 'Margin of Error' with 'Margin of Votes".
In all polls, the 'Margin of Error' applies to BOTH (or all) parties - and predicts the final margin one can expect EITHER alternative to amass to within a certain very high percentage of accuracy (see the post just above this for the details).
However, if a certain percentage of voters CHANGES THEIR VOTE, IT IS AT THAT POINT that you 'double' the final result. In other words, if OBAMA had a 2% lead over Clinton and 1% of his supporters changed their minds, THEN you would have to 'double' the percentage change, since the former 2% lead would now vanish and they would be tied (since Obama LOST 1% while Clinton GAINED 1%.)
May 2, 2008 2:41 AM | Reply | Permalink
Eric=Taylor Marsh
Gets more like her everyday...Sad isn't it!
May 2, 2008 3:25 AM | Reply | Permalink
Actually, Eric is essentially correct in his interpretation of the margin of error. It applies to each individual result in the poll.
With a margin of error of 3.9, and the assumption that the poll's sample is random and representative of the population being sampled, there's a ninety-five percent probability that the Obama's actul percentage of the vote will fall between 46.1 and 53.9 percent ninety-five times.
With the same assumptions, (random and representative sample), Hillary's vote has a 95 percent probability of falling between 40.1% and 47.9%.
So, yeah, the results of the poll do show the lead within margin of error.
May 2, 2008 4:34 AM | Reply | Permalink
There's a lot of misinformation going on here. I'm an engineering grad student right now, so I'll be happy to chime in.
The margin of error represents the standard deviation on a bell curve, normal distribution. The proper way to read this poll for Clinton is as follows. Clinton has 42% of the vote with a margin of error of 4%. That means that there is a 68% chance that Clinton's "true" vote % is between 38 and 46 points. This is because 68% of the total curve is within 1 standard deviation of the average. 95% of the normal distribution curve is within 2 standard deviations, so there's a 95% chance that Hillary's "true" vote is betwen 34 and 50 percentage points. You can calculate this the exact same way for Obama, using the same standard deviation but using his 50% poll number vote.
So that explains the correct way to read the poll. However, it doesn't answer the REAL question, which is what is the percent likelihood that Obama is 'truly' ahead of Clinton. To calculate this, you must calculate the standard error using the 50%, 42%, and n=650. Then calculate the z-score and then look up the z-score on a statistics table. I just did this in Excel and got that for this poll the z-score is 2.13, which corresponds to 98.34%. Therefore, this poll essentially states that they are 98.34% sure that Obama is 'truly' ahead of Clinton.
Of course, this should not be interpreted as a 98.34% chance that Obama will win Oregon, since the primary was not held today and things can happen (i.e. idiot media pundits looping Wright) that can cause the numbers to change. But the poll results show that the day it was measured, there is a 98.34% chance that Obama is 'truly' ahead of Oregon.
May 2, 2008 5:12 AM | Reply | Permalink
Except that margin of error indicates a 95% confidence interval, not just one standard deviation.
May 2, 2008 8:18 AM | Reply | Permalink
There's a wrinkle in that there must be some negative correlation between the percentage support for Obama and the percentage support for Clinton, given that they're the only two candidates and there aren't that many undecideds.
To illustrate this point, Obama could run as high as 54% within the 95% confidence interval, while Clinton could run as high as 48%, but these figures add up to more than more than 100%. The chances of that happening are zero.
May 2, 2008 8:38 AM | Reply | Permalink
Sorry for the awkward typos in my post above. The basic statistical fact is that if the sample is random and representative of the population, the margin of error must be added and subtracted from each statistic to obtain the range within which the actual result will fall with a probability of 95 percent. So:
Obama 50% (95% probability that his result will be between 47.1% and 53.9%)
Hillary 44% (95 percent probability that her actual result will be between 40.1% and 47.9%).
The chances of any particular result within that range are theoretically related to the confidence level. So, there's a 90% probability that the actual results (if the whole population were asked who they preferred) will fall within about 2.2% of each of the poll's results.
Because of the number of undecideds in this poll, it can't be considered stable. Undecideds range from 2.1% to 9.9%, a range of over 400%.
May 2, 2008 4:59 AM | Reply | Permalink
"The basic statistical fact is that if the sample is random and representative of the population, the margin of error must be added and subtracted from each statistic to obtain the range within which the actual result will fall with a probability of 95 percent."
Perhaps in your alternate universe, but not on this planet. In a normal distribution, 68.3% of the probability falls within plus/minus 1 standard deviation of a SINGLE poll number, 95.5% within +/- 2 standard deviations, and 99.7% falls within +/- 3 standard deviations. This applies to EITHER Obama or Hillary's number, but NOT the difference between the two numbers. Think of it as a question of two overlapping bell curves, asking what percent overlap there is. You must calculate the z-score in order to find the percent probability that Obama is leading Clinton. Any analysis without a z-score is faulty.
May 2, 2008 5:20 AM | Reply | Permalink
L0ngT0m is correct.
The margin of error represents a 95% confidence interval, not a single standard deviation, and it applies to both figures, not the difference between them.
May 2, 2008 8:27 AM | Reply | Permalink
Oops, I thought that the poll was Hillary 42%, but it's actually 44%. Recalculating the z-score I get 1.58, which correlates to 94.3% that Obama is 'truly' ahead of Hillary.
May 2, 2008 5:15 AM | Reply | Permalink
The moral of the story is that the concept of 'you must be ahead by 1 (or 2) margins of error for it to count' is ill-founded and only being pushed because most political pundits have never taken a statistics class. The REAL question is always what percent chance is Obama truly ahead of Hillary. Here it's a 94.3% chance, so most people will qualitatively express that as a definite that Obama is ahead. Generally speaking, a lead of at least 2 margins of error will put you somewhere around 97% or more chance of being right about who is 'truly' leading.
May 2, 2008 5:25 AM | Reply | Permalink
You're wrong. It's the range that the margin of error provides that's normally distributed about the poll result. If you did the poll a hundred times, 95 of those times Obama's result would be between 47.1 and 53.9, and the frequency of the results oughtta be a normal distribution around 50.
So, while there's a 95 percent chance that his actual result would fall between 47.1 and 53.9, the probability of the result being 50 is greater than the probability of a result being 47.5.
Your statement that "Here it's a 94.3% chance [that Obama is truly ahead], so most people will qualitatively express that as a definite that Obama is ahead." shows that you sort of understand that the poll results are within the margin of error. If they weren't, the chance that Obama was ahead would be 100% (assuming random representative sample etc.)
May 2, 2008 10:05 AM | Reply | Permalink
Trust me, I'm right. Engineering grad students deal with statistics on a nearly daily basis (or so it seems). I noticed you didn't even bother to calculate a z-score with your analysis. Do you even know what that is? This is like the 8th lecture in an intro to statistics course in undergrad. I've been a math whiz since elementary school, so please come correct. :-)
By the way, if you don't believe me, just look up margin of error on wikipedia.org and they'll give you the outline as to how to approach this question.
One thing that does deserve mentioning though is that statistical analysis of polls assumes complete negative correlation of results. Interpret that as all of the undecideds split equally between the two candidates. So that's an assumption that goes into calculating the percent likeihood that Obama is 'truly' leading Hillary.
May 2, 2008 1:05 PM | Reply | Permalink
No poll would ever have 100% assurance, unless it polled EVERY SINGLE VOTER in the state. There will always be a margin of error with polls since they poll hundreds to a thousand people when hundreds of thousands of people will vote.
May 2, 2008 1:13 PM | Reply | Permalink
Zogby or TeleResearch?
May 2, 2008 7:40 AM | Reply | Permalink
Can't trust a pollster who doesn't understand the difference between correlation and cause and effect.
May 2, 2008 7:56 AM | Reply | Permalink
Too True.
It doesn't appear they asked a question about Wright so it is a bit of an irresponsible statement. It is likely true, but people in his biz should know better and there is a difference between hypotheses and fact.
May 2, 2008 8:18 AM | Reply | Permalink
I am sure other polls will confirm the trend to Clinton in Oregon. I think that she will gain and will likely have an excellent opportunity to win. Having grown up throughout the Northwest myself, previously even lived in Portland, I can say that while there is a core liberal base, it is still quite conservative and moderate in some respects. Further, this is not the South or big eastern city environments. Growing up out here we did not learn the racism that so often clouds perceptions elsewhere (yes, because there are so few Blacks). What this means is that folks will be more shocked and turned off by Obama's 20 years with Wright and all of the things that have come out than others might back east. Further, except for a small, radical group in Portland, the matter of hating America will be received very poorly.
Matthew
May 2, 2008 8:05 AM | Reply | Permalink
Geaux Hokies!
BS, '99
May 2, 2008 8:29 AM | Reply | Permalink
The Zogby poll was conducted Wed & Thurs so it is the first one conducted complely after Obama's denunciation of Wright.
I don't believe Obama will do worse in IN than he did in PA so I don't believe the Teleresearch poll. I am not sure I belive Zogby either but I think Zogby is closer. The nice thing about Zogby is this is the first installment of a tracking poll, so even if they are off on the actual number, they should capture movement fairly effectively.
May 2, 2008 8:32 AM | Reply | Permalink
I notice that this post is bereft of expectations game spin about who "needs" a big win. Isn't it the case that Hillary needs to win every remaining primary by a wide margin if she is to capture a delegate lead going into the convention? I've heard that she needs %71 of remaining pledged delegates.
May 2, 2008 8:42 AM | Reply | Permalink
Another DNC Chair Paul Kirk kicked in today.
http://www.dailykos.com/storyonly/2008/5/2/83155/85648/878/507640
Sorry, Greg & Eric.
May 2, 2008 8:56 AM | Reply | Permalink
BREAKING: I just received a text message from Howard Wolfson. Oregon does not have a lot of poor Appalachian whites. Therefor it does not count.
May 2, 2008 8:59 AM | Reply | Permalink
THE STORY THAT IS ABOUT TO BLOW
Clinton campain talking about Indiana voters...
This is the link to the Indiana trashing of voters by Mickey Kantor-Clinton staffer
So typical of the Clintons.
Can you say...Sh!tgate!!
http://www.youtube.com/watch?v=e-MzByUHIzw&feature=email
May 2, 2008 9:56 AM | Reply | Permalink
THE STORY THAT IS ABOUT TO BLOW
Clinton campain talking about Indiana voters...
This is the link to the Indiana trashing of voters by Mickey Kantor-Clinton staffer
So typical of the Clintons.
Can you say...Sh!tgate!!
EVEN WORST THE ENTIRE CLIP:
http://www.youtube.com/watch?v=sN_nQOHj__s
May 2, 2008 10:01 AM | Reply | Permalink
FYI--It's a little off to say the Oregon primary is May 20. Oregon is vote by mail. The ballots go out May 2-6 and people have til 5/20 to get them in.
May 2, 2008 11:04 AM | Reply | Permalink
Back in 2004 some math/stat profs gave Kevin Drum a neat little 5-line Excel spreadsheet that you can use to calculate the probability that the poll results reflect an actual lead (or you can use the table in his column to estimate):
http://www.washingtonmonthly.com/archives/individual/2004_08/004536.php
According to that spreadsheet, the probability that the poll results above reflect an actual lead for Obama is 94.3%, which is good enough for me (and Hokie). :-)
As for Margin of Error - Eric, I do not think that term means what you think it means. But I applaud you for reminding your readers that the uncertainty exists.
May 2, 2008 11:12 AM | Reply | Permalink
That table is essentially an abbreviated z-score table adapted for polls. Great to use. Political pundits should use it. It's easy to understand even without a background in statistics.
May 2, 2008 1:23 PM | Reply | Permalink
Yup, here in OR we get our ballots any day now and start voting by mail. There will be a bunch turned in during the first few days, then a slow trickle until the last couple days before May 20 when the bulk will be turned in. My uneducated guess would be the "hi-info" type voters tend to get it done first (you have to really study and we have a lot on the ballot) and the low-info voters will vote last minute.
So-- me and my hubby will be out canvassing for Obama this weekend.
I've seen literally one Hillary window sign and a bumpersticker. Lots and lots of Obama. Portland is Obama country. I haven't been out in the 'burbs or country lately to see how things are shaping up there, however.
Hoping that after May 6 Obama comes back and does a major bus tour around our beautiful state.
May 2, 2008 11:23 AM | Reply | Permalink
Oh and one more thing-- Hillary's picture in the Oregon voters pamphlet is lovely, but I swear to god it's airbrushed. She looks about 35. Having run for local office myself last year I know they require a RECENT unretouched photo. Hmmmmm.
May 2, 2008 11:29 AM | Reply | Permalink