Gallup's daily national tracking poll shows a three day tightening in favor of Sen. Barack Obama. Yes, it's over a weekend, so we'll have to wait a few days to see whether it sustains itself.
« Obama Slams Clinton Veep Illogic | Main | Two More Superdelegates For Obama » Gallup Retightens In Obama's Favor10 Mar 2008 01:26 pm Comments (30)
How is going from a tie to a five point lead a tightening? So basically, the state of the race is the same as it has been the last couple of weeks, with a little blip in there that moved the polls a couple of points when the kitchen sink hit just in time for Clinton to score a couple of marginal wins. This thing is so over. 
Gallup Retightens In Obama's Favor Marc, look again at the graph. Over the last three days, the tracking poll spent one day "retightening," and then two days reversing and expanding in Obama's favor. So, no, it basically doesn't at all do what you say it says. Not that national tracking polls particularly matter at this stage of the game, but I swear, you must throw some of this shit up without a moment's thought. Could you please go down the hall to Fallows's office for a quick refresher on communicating clearly and precisely in the English language? If he's in China right now, I order you to wait there until he gets back. Also, I thought the party line was that Hillary does better on the weekends, since the lazy college students who are Obama's sole base of support are out getting drunk and hooking up, and in any event only carry cell phones anyway. Clarification, please!
WHY CAN'T CLINTON CLOSE THE DEAL?!?!??!
Wow. If this race "tightens" any more, Obama will be up by 10!
Negative campaigning works...but for how long and at what price???
I think he just means that Obama re-tightened up his support. It's not the most elegant way to put it, but I don't think he's mischaracterizing anything. The real question this poll raises is: how the heck is Mike Gravel getting 2% of the votes? Even if it's 1.5% rounding up to 2%, that's still 19 people. I can't believe there are 19 Gravel supporters in any random cohort of 1,300 people. Then again, he did get over 6,000 votes in California.
I guess Mark Penn was right about stopping his momentum.
I, too, am confused by the use of the word "retighten." Hillary had a lead of 4. I could see "retighten" in reference to the shift to what looks like a tie at 45 (a 0 point gap is tighter than a 4 point gap). Or even in reference to a 47-45 Obama lead (a 2 point lead for Obama is smaller than a 4 point lead for Hillary). But how is either local movement (from a 2-point to a 5-point Obama lead) or broader movement (from a 4 point Hillary lead to a 5-point Obama lead) "retightening?" Marc- I like so much of what you do. Could you please rewrite the title from its current Clintonian language? I'd hate to see the phrase "it depends on what the meaning of retighten is" creep into common parlance. Thanks.
Kind of like Hillary's hoochie tightened after Chelsea popped out.
Who are these 4% of the voters who are changing their minds day to day?
No doubt about it, this race is tight. After Obama clinches the pledged delegate lead, it'll get even tighter. I expect we'll see a maximum level of tightness - peak tautness, if you will - right after the convention. Jeebus, reading this blog makes my brain hurt
How could this be? When Hillary was in China, she gave a speech! A speech. She crossed the threshold.
Oh boy. People here are certifiably insane. Despite was Gallup says a 5% lead over one day in a rolling poll is NOT statistically significant. What that poll, along with the Rasmussen tracking poll, tells us is that the race is very close and volatile at the national level. That's the reasonable interpretation. Not supposition about who is leading on a single day or whether it's tightening or widening.
People here are certifiably insane. I kinda like it here. My preference is for my nutjobs to be of the literate variety.
Tim, I don't think you should be casting aspersions simply because others disagree with your own favored interpretation, which, frankly, I find to be as strained as Marc's. 1. Marc's formulation was, at best, a very poor choice of words. It just doesn't fit with the traditional and normal use of those terms. For the very reason that I like to see the English language retain its significance, I take umbrage with his phrasing. 2. The Gallup figure is drawn from three days of responses; Obama's lead is not the product of a single day as you suggest. 3. The Gallup poll has a +/- 3 MoE. Therefore the five-point lead is statistically significant, as were HRC's four-point leads last week. 4. Similarly, the swing from +4 HRC to +5 Obama is statistically significant, and most likely not only a product of a "very close and volatile race." 5. The Rasmussen poll is not relevant to, and proves or disproves nothing with respect to, the methodology or statistical significance of the Gallup poll.
Tim K- But, per Gallup itself: "giving Obama a slight but statistically significant five percentage point lead." So I believe the "NOT statistically significant" statement that you made to be untrue. What I'm trying to interpret is use of the word "retighten." It makes no sense to me, and Marc's pretty literate, so I'm hoping he will explain.
Jeff Larson: Yes the 49-45% "lead" is the product of a single day given that Clinton was leading 47-45% yesterday. That means this shift is the result of one evening's polling, namely Sunday's. A 5-point lead is not really statistically significant on a +/-%3 margin of error. Why? Because the margin for error is 3% for each candidate, but the margin for error is double that for determining the difference between two candidates levels of support. If you don't believe me why don't you input the information into American Research Group's "Ballot Lead Calculator"? Here is the link: Barack Obama's support could be anywhere from 46%-52%, and Hillary Clinton's support could be anywhere from 41%-47%. So Obama could be ahead by 11% or behind by 1%... we really don't know. Only time will tell.
1. You interpret the Gallup graph incorrectly. Yesterday Obama led 47-45%. 2. Even if Clinton led yesterday, you cannot prove with the information provided that this is the result of a single day's polling. It is possible (and equally unprovable) that Hillary led massively on Day 1; led modestly on Day 2; trailed modestly on Day 3; trailed modestly on Day 4. With the right values for each day, you can have statistically significant swings that resulted in a Clinton lead until today, but that is not the result of a single day's polling (at least as you seem to mean it). 3. Like Tony above, I will refer you to Gallup's statement that today's polling represents a statistically significant lead. I suspect they probably know better than you, me, or you and me combined. 4. I will read and evaluate your link, but I don't have a lot of respect for ARG or their work. I'm much more inclined to regard Gallup as a credible source.
Looking again at the Gallup link, I see that they actually state that the result in today's poll is the result of three days of Obama holding a lead. So unless I misunderstand you, no, this is not the result of a single day's polling (at least not in any meaningful sense). I'm looking at the formulas behind ARG's calculator, and I'm not yet clear on why they made a couple assumptions they made. I'll try to teach myself enough about statistics to come up with some kind of meaningful response here, but in the meantime, I see no good reason not to take Gallup at their word.
Jeff Larson: Clinton didn't have a statistically significant lead over Barack last week either. Without Gallup releasing each day's individual numbers we cannot know whether there was a statistically significant lead on any single day. Although it's also worth noting a single day's sample is only about 300 voters, so that has an even wider margin for error (probably more like 6-7%). ARG has had it's share of embarrassing misses this cycle like in Iowa, just as most other polling firms have. ARG was very accurate in Ohio and Texas, incidentally. Gallup is a very respectable polling outfit, but I that doesn't mean the person who wrote that particular post didn't make a mistake. It wasn't necessarily Frank Newport's statement. I don't claim to be a polling expert, and never would, but I'm pretty sure I'm correct about this.
I took a half dozen stats courses in grad school, and convention in my discipline is to think of 2 SE's (rounding a bit....much depends on distributions....) difference as statistically significant. If Tim K's point is that 5 points doesn't hit 2 SE's, he might be right. He might also be wrong. I suspect that the Gallup people know what they are talking about, and Tim hasn't addressed the quote from Gallup that I included above. I also note that there is rounding here. Is the lead exactly 5%? Probably not. Is the SE exactly 3%? Perhaps not. Suppose the lead is actually 5.5% (49.25 to 43.75) and the SE is actually 2.75%, each rounding. In that case the lead is at 2 SE's, which would be consistent with Gallup's statement. But all of this is, to my mind, beside the point. Marc has talked about the race "retightening." And I have no clue what he was thinking in his use of this word. I do not know the time frame of which he speaks. And I find it weird for him to be talking of it tightening when Obama's lead is larger (albeit not by a significant amount) than Hillary's lead has been at any point.
I didn't see Tim K's 5:09 in my last reply. It is certainly possible that someone writing up the Gallup story erred. I still don't know why Marc used the word, though.
Tony: Thanks for your insights. I'll defer to your much more precise knowledge on issues concerning standard errors and confidence intervals. However, to my mind, a 5% lead does is not very useful in terms of knowing who is ahead. Perhaps it is just barely statistically significant, or not quite, either way it's not *politically* significant. We've seen a lot of volatility with these daily numbers.
Tim- You're welcome. It's certainly hard to know what's politically significant these days. My own take would be that the poll's political significance is that any bump that Hillary got coming out of last Tuesday has melted away. But I'm an Obama supporter, so have been hoping for that effect and it will help me sleep a bit better these next couple nights. Sounding like a broken record....The headline still strikes me as very weird. How to construe this as tightening is just very, very unclear to me....And that's what prompted my first post.
Tony: Well you may be an Obama supporter but you seem as if you are open-minded. I tend to agree with you that it looks like Clinton's small bump after Texas and Ohio has dissipated. But I was equally struck by how the slight lead Obama built by winning 11 consecutive contests in February evaporated even before Texas and Ohio had voted. Neither candidate has a clear advantage among Democratic primary voters at this point. You're right about Marc's confused language. How do you see the races playing out moving forward?
I think this proves my theory right that Samantha Powers calling Hillary a Monster actually helped Obama. A reasonable insult coming from an accomplished white woman that was echo chambered by the media actually drew a little blood....it was friday night that gallup's track started trending Obama, which was right after the Powers story broke.
RKA: How could calling somebody a "monster" draw any blood? It's just the kind of over-the-top remark that can only embarrass your own side. The idea that her remark triggered a change in the polls in Obama's favor is frankly too stupid for words.
Right after that statement, the narrative changed from HIllary's comeback to HIllary's bad behavior. Low infomation voters are susceptible to this stuff Tim K, they don't all think like blog posters like us. Clinton uses these techniques all the time...i was impressed that Obama was able to coopt and perfect the Bob Johnson method of hitting your opponent with far more precision than Hillary ever could. The whole thing was brilliantly executed. Bravo, Bravo!
Tim- Thanks. I try to be openminded, but I'm sure I'm not in some ways. As to the moves...As I understand these things, they do rolling averages. So each day, they add one day's data and drop another. Obama peaked on Saturday, March 1 at 50, with a corresponding trough for Hillary at 42. Obama's average dropped each of the next days, as polling was added from Sunday, Monday, and Tuesday. (And there's a mirror Hillary rise those days.) Hmmmm....speaking late in the evening after a quick look, seems like things stabilized on Wednesday the 5th. Whatever sample came in matched the one from Sunday the 2nd that would have been dropped to add data from the 5th. So maybe I was wrong about a bump. As you point out, the movement toward Hillary was pre-Tuesday. The phone, Nafta-gate, and such? Whatever, it seems like it's gone back to where it was. What that means about depth of support...dunno. I wish I did. I read blogs trying to sort it all out, but haven't gotten a good handle. How does it play out? I just read a NYT piece suggesting that Hillary is now playing to the superdelegates. Can she convince them that Obama won't win in November? If she does, she could get them to come aboard. If I'm Hillary and my goal is to have myself become president this year, that strikes me as a rational decision. She can't get the pledged delegates to win. But man, that's one risky strategy for the party as a whole. Will her attacks on Obama undermine him enough to weaken him, but not enough to get her elected? If it does, that could be McCain's path to the White House. To be honest, I'm hoping that her strategy backfires, and that a fair number of voters move away from her to Obama, giving more supers cover to declare for Obama. I saw at dkos that about 13 supers have declared for Obama in the last week to something like 4 for Hillary. A stampede and a PA win could finish it in April in time for him to pivot to contrasting with McCain, while Obama has the money advantage. If I'm a Hillary supporter, I'd think my hope would be seeing Obama's numbers drop back to where they were middle of last week, and a bit more, and hope for something of a reverse stampede. But I don't know what changes the game for her between now and April. Does she have new attacks up her sleeve? I'd think she'd have used all her ammo to get Texas in the win column. If she doesn't change the game with new attacks, I'd think Obama would win out. We'll see. Bedtime for me. Ooops...forgot one PS. When I looked at the Gallup report, I saw that they said the "maximum" error was 3%. Error depends on sample size, and they report having no fewer than 1000 in each sample. In the particular sample (March 7-9), they had more than 1000....1297. There are diminishing returns as you add to samples, so it's not like adding 300 voters reduces error by 30%. But I think it's safe to say that the margin of error will be a bit under 3% in this case.
|
Subscribe to this blog by Email Twitter || On Facebook || QIK Live Channel  Recent Posts
Monthly Archives
|
| 
Copyright © 2008 by The Atlantic Monthly Group. All rights reserved.
I don't understand. Could it be that Hillary overreached in her criticisms of Obama? Could it be that all that talk of crossing the C-I-C threshold is coming back to bite her in the ass?
How could that be? Clearly the world now knows that Hillary and I solved the crisis in Kosovo!
Clinton/Sinbad '08 - We'll make a third-party run if we have to you, ungrateful bastards!
Posted by Sinbad | March 10, 2008 1:34 PM