Is Obama's Lead In Gallup Polls Stable Or Volatile?

If your heart stops with every dip and bump registered by Gallup's daily tracking poll, here's some welcome news: the race is more stable than you might think.

Emory University Political Science Professor Alan Abramowitz sends along a couple of interesting graphs that each plot Barack Obama's consistent lead over John McCain in the presidential race. In the first graph, each daily tracking poll since the end of the Democratic primary season is charted as a distinct data point, making for what looks like a wild ride:

Having some fun with those who would over-interpret this graph, Abramowitz said: "Look--Obama is surging! OMG, Obama's support is collapsing. No wait, he's surging again! Oh no, he's collapsing again! What happened to that 9 point lead? McCain's strategy must be working. All is lost--the end is near! I knew we should have nominated Hillary!"

Then Abramowitz steps back a bit, averaging the Obama lead from each daily tracking poll into 10-day data points, stretching from June 5 to August 4. Which gives him a rather stable-looking 3-to-4 point lead for Obama:

Explaining the optical illusion, Abramowitz says: "When you group the tracking poll data into ten-day periods it becomes clear that support for the presidential candidates since the end of the primaries has been extremely stable, with Obama maintaining a modest but consistent lead, and that almost all of the day-to-day movement in the data is simply random noise."

Or, in a somewhat shorter version of his attitude toward the ups and downs of the daily graph that gets turned over each day on cable news shows, Abramowitz says: "Never mind."

Comments

[JJeff88]

The article makes a good point. Consider these other things:

You can make the tracking look more volatile by increasing the inches between the numbers down the left side. Conversely, if you want to make the trend look more stable, just narrow the gap between those numbers.

Regardless of how you flatten out the trend line by looking at it in terms of weeks instead of days, the fact that there's that much day to day volatility makes the whole polling process suspect. As Chuck Todd pointed out to a Gallup representative yesterday: It's a tough sell to convince me that 10 million people are changing their minds back and forth on a daily basis. This suggests that either the sample size isn't big enough, the questioning process may be flawed or the data may have been tweaked by humans in order to compensate for perceived sampling or questioning weaknesses.

What's not being covered by the media (probably because it's too wonkish and difficult to explain in a 30 second byte) is the "inside the numbers" story - i.e. How was the sample size and composition determined? How were the questions worded & in what order? Are the sample numbers "pure" or were they massaged in order to balance out the demographics? In other words, to what degree was the poll a "let the numbers fall were they may" deal? And to what degree did human judgment (with all its frailties) enter into sampling, questioning and analysis?

[BeatlesMania4]

This is exactly my problem with polling in general, thank you so much for voicing it so clearly and concisely. Human judgment inevitably enters the calculations to influence what the pollers think should be a representative sample.

I'd like to see how these numbers react to specific incidents however and more about how certain people that are tracked react to different incidents throughout the campaign. I'm particularly interested in how the most recent scandal thats being rumored affects his numbers: http://tinyurl.com/6jb7l6

[rollingdivision]

Democrats have an 18 percentage point lead in generic polling and their presidential candidate is tied with his Republican opponent. Any objective observer would conclude there must be a problem, but don't worry.

[decisivemoment]

By any objective statistical standard, Obama has a lead. You don't take 55 polls, ALL of which Obama leads, and say it is tied. Statistical significance is acknowledged at a 95 percent level. You'd want let's say, three of those 55 to show a McCain lead to call a tie. McCain is not even tied in any of them.

That said, Obama ought to be leading by more. And in that sense, there's a problem.

[rollingdivision]

"Statistical significance is acknowledged at a 95 percent level", please, by what statistical analysis do you calculate this number?

[stevievi]

Decisivemoment
You have to remember who's being polled. I'm 32 years old and don't have a land line. I haven't had one in maybe 10 years. Most of my counterparts are in the same boat. Now our parents have land lines and their parents have land lines. So it's obvious to me and 95% of my friends and co-workers who don't have children don't have land lines so were are not being polled.

The polls are very insignificant when you bring that into equalization. It's seems to me that the same type of people are being polled- older white Americans. Who really at this point doesn't know that he's Christian (20 years same church) and not a Muslim- so the fact that these polls keep saying that 12% of the people polled think he is. Those people are not living in a vacuum, they have friends and family who know the truth.
I have family of many colors all over the us and the Caribbean and not one of them have been polled. Not one.

[Bluesman48]

Two problems, actually. Racism and stupidity.

[huffposeur]

i agree. it's the stupidity of obama supporters who think that anyone who doesn't drink the kool-aid is a racist.

although it's difficult for people here to understand, it's possible to look at obama and not see the color of his skin but the thinness of his resume.

it is possible to not like obama, not support his positions and NOT be a racist.

[rasthedestroyer]

you forgot one... fickle democratic base.

[NeoStar9]

This is what happens when you have a country where a large part vote against their own self-interest or are either stupid (lets face it there area lot of stupid and gullible people in the USA). Then there is the racist factor that has to be accounted for. No matter what the policy is there will be a group of people (male and female) that won't vote for a woman or a black person or hispanic or even asian person no matter what. This is why things seem close. Also we have no clue as to how these polls are being conducted or what the actual questions being asked are.

That's the huge thing. What questions are being asked. Are they straight yes and no questions? Are they slanted one way or another? Are people allowed to interpret them and then put them into a column based on that?

The way the media presents things I think all national polls are in question. State ones tend to be more important I think since elections are won on a state level. However I don't care if Obama wins by 10 points or by a single vote that causes him to get that last remaining state that will tip the balance. As long as he wins.

[uvymopkq]

Correct...they mean nothing. Because he continues to shoot himself in the foot.

[Nonamnesiac]

In 1986 or 1988 (I forget which year) NJ Senator Frank Lautenberg was running against General Pete Dawkins (an Army football phenom in his youth, kind of like Staubach). At the end of September of the election year, Lautenberg was 17 points behind in key polls and won the campaign anyway.

This daily tracking poll BS is simply designed to keep the army of pundits on the 24 hour a day commentary stations (they are no longer news stations) chattering stupidly, inaccurately and endlessly, attempting to shape the opinion of the electorate and limit debate so that progressive solutions are seen too radical while moving the political debate to the right.

These are polls taken even before the political conventions have occurred. I can't believe they're given such weight by serious people, especiallty given Hillary's inevitability and the numerous other outright polling inaccuracies that have been exposed during this never-ending political season.

[MNmommy]

The cell phone argument is interesting to be sure, but this is the first time in eight presidential election cycles as a registered voter that I've ever been polled. Rassmusen called, I about fell out of my chair. I always feel like getting polled is sort of like jury duty, rarer than it should be.

[bellecloche]

The other point is that - like it or hate it - the president doesn't win by popular vote, he wins by racking up states, so national polls mean very little. I'm grateful to HuffPo and DailyKos for following the state polls because THEY matter, and much more so than in the primaries because the electoral votes are not proportionately allocated. The print media rarely reports on those.

[perhapshowever]

Popular vote and polls that attempt to measure across the entire country matter not.

It is all about the electoral college and individual states.

Obama is in deep trouble, he does not seem to be able to pull in independent voters like myself who have had enough of his wind vein tricks.

Bob Barr is real change and I've had enough of this two party system.

[research]

Land line phone polls are pure BS. remember how bad the primary polls were.

[goodog]

Now, the polls the McCain campaign doesn't want you to see:

http://www.pollster.com/polls/us/08-us-pres-ge-mvoand3s.php

When looking at the two-man race, we see McCain gaining, and we assume he's taking away from Obama. But you might notice, Obama is gaining too. So what gives? How can they both be gaining. Someone has to lose.

It's Bob Barr, the Libertarian candidate who's losing. As he ticks down, McCain ticks up. Their in-synch. McCain's gain is simply a consequence of Libertarians realizing they don't like Bob Barr, who is, after all, no Ron Paul.

About the poll:

It's true, you can't trust any one poll, which is why it's necessary to check these consolidated average polls.

http://www.realclearpolitics.com/epolls/2008/president/us/general_election_mccain_vs_obama-225.html#chart

http://www.pollster.com/polls/us/08-us-pres-ge-mvo.php

Consolidated polls, though little reported, have accurately measured public perceptions and preference through the primary. The problem is, their stability doesn't drive the news day.

"But... but... people will say anything. Some lie to pollsters on purpose!"

Which is what consolidated polls are all about. The average of several polls spreads the risk of anomalies, bad data, and faulty processing across all polls, mitigating error.

[whizkid]

But will the polls show when the truth comes out post-convention?

http://www.thecityedition.com/Pages/Archive/Winter08/2008Election.html

[UrbanRevolution]

Look whizkid, Obama did not STEAL the election nor was he "handed" the election by party insiders. Superdelgates have always been able to vote pre-convention. No one said boo when Clinton had the lead before the voting by regular folks already started because she had 233 superdelegates. Now that she is an Obama supporter as opposed to opponent - the superdelegate votes are an issue? I don't remember you nor anyone else complaining about the superdelegates on Jan 1st.

[EvoMan]

Still haven't gotten over it yet, huh?

By the way, you've linked to a load of malarkey.

[whizkid]

Evo
Malarkey? Or just unpleasant reality?

[HowietheScreamer]

Yes, of course, but the MSM has a vested interest in the wild swing story.. it makes more money

[PumaAnn]

What wild swing? He's essentially tied and remains stuck. 2.7 RCP ........ that's tied.

I am so not grasping this story.

O tanked in Feb. He really never has come back.

[AdamND]

He won.. get over it.

[jorydozier]

I don't believe winning the most votes, the most states and the most delegates can be considered "tanking".

[decisivemoment]

Tanked in February? His month of ten wins?

Let's at least get the historical details correct before performing analysis.

And 55 polls, all of them showing an Obama lead, does not constitute a tie.

I'll concede that a 3.5 percent average in the tracker constitutes an inadequate lead under the circumstances, but do not forget who the Democrats are running against -- the biggest media darling ever, the "maverick", the one who gets sprinkles on his donuts from the AP, the one who can get six months of good press from one barbecued chicken wing, the one who the yellow-headed misogynist himself on MSNBC acknowledges that "we're his base."

Obama's main unfulfilled challenge is to define McCain for what he really is, not what the media say he is.

[rasthedestroyer]

ok little kittykat your 9 lives are up ..now go away

[lthuedk 1]

Similar in more than name.

http://www.light-to-dark.com/ann_coulter_republican.html

[dumboldfarmer]

The polls mean very little at this point. We haven't even had the conventions yet.

[Oldtt]

Both candidates are presumptive nominees, so the polls mean about as much now as they will after the convention. Right now they show a close contest in a year that should be a Dem landslide. I guarantee this is not lost on the candidates, and will affect VP selections and allocation of money to toss-up states. The fact that O's world tour did not result in a significant bump is a problem for Dems.

[ChimpNchief]

And, The fact that McSame has NEVER had a lead in ANY poll is a problem for Rethugs.

[djelimon]

Why? They actually said they didn't expect a bump at all while on tour. Then there actually was a bump for a while.

In terms of electoral votes O is doing quite well.

[Bagger]

President McCain.....think about it.

Now get to work! Raise money, volunteer, get out the vote! Our lives depend on it.

Obama says my mama.

[Mildmannered]

And think what damage a President McCain can do to our Supreme Court.

[GHENT007]

The media is doing everything that it can to make mccain look viable, so they put out fake ass poll numbers to make the race look close we all know that barak obama has a good 10 to 12 point lead over mccain but they think they can get the suckers to vote for mccain because he appears to be close! sorry!! republican media monkeys!! the democratic majority sees through your manliputation of the polls and were gonna HAND YOU YOUR ASS IN NOVEMBER!!!!

[Texas4Obama]

The majority of the people that will be voting in November have NOT been polled.
All the people that don't answer their phone when solicitors call haven't been polled.
All the people with cell phones are not being polled.
Polls go up and down and don't mean anything - the only thing that counts is the results of the November election.

And if you do believe the polls then you have to realize that McCain is NOT ahead in the polls.

Obama '08

[BetterDays]

Since Obama brings the youth vote, the polls may underestimate his lead. The youth he has attracted typically don't have the land lines used for polling. It is possible that there are corrections in the polls for this dynamic but not certain.

[dajay]

d