# Pollster.com

## Articles and Analysis

### POLL: USAToday/Gallup National (7/25-27)

USA Today/Gallup
7/25-27/08; 791 LV, 4%; 900 RV, 4%
Mode: Live Telephone Interviews
(blog entry, story, results,
"Assessing the Impact of Obama's Trip",
"Who Are Likely Voters and When Do They Matter?")

National
LV - McCain 49, Obama 45 (June: Obama 50, McCain 44)
RV - Obama 47, McCain 44 (June: Obama 48, McCain 42)

Alan Abramowitz:

How do you get from a 47-44 Obama lead among RVs to a 49-45 McCain lead among LVs?

A few quick calculations shows how. You have 900 RVs and 791 LVs, so that means that among your 109 UVs (that's unlikely voters according to Gallup) Obama leads McCain by a whopping 61% to 7%.

Putting it another way, according to Gallup 16% of registered Obama supporters are unlikely to vote compared with only 2% of registered McCain supporters.

____________________

jsh1120:

Prof Abramowitz,

Does this mean that Gallup is using a simplistic binary variable to classify respondents as LV's? I would have expected a probability weight. Or perhaps it doesn't matter?

____________________

1magine:

It means, if you are 18-25 and have not voted before, if you are black and do not vote often, or if you are Democratic and your party has a history of not coming out to vote - - this poll has found that you are unlikely to vote this time out.

Clearly - this is the case? Surely everyone must see that if the electioon were held today it would be JM in a landslide electoral college and popular vote victory. Why - I'm personally going to remove my voter registration and not even bother to follow the race any more. It's over and somehow BO lost. Ohh well. TG we have USA today to properly weight the Gallup poll, unlike that messy daily tracker that somehow surveyed 3 times the sample came up with similar raw numbers and improperly weighted the race as Obama leading McCain. When we all know Obama isn't leading...

____________________

carl29:

I think this poll is kind of "crazy" not only for the numbers but who is putting it out, Gallup. Today's gallup tracking poll reflects the presidential preference of a sample of 2,674 registered voters from 25-27 of July, the exact same days the poll with USA Today.
Gallup tracking poll, which has a bigger sample and a 2% margin of error, shows Obama leading McCain 8% among registered voters. However, its poll with USA today, which has a smaller sample of registered voters and a 4% margin of error, shows Obama leading by 3%. The disparity is pretty obvious and that makes it a little more "weird" coming from the same pollster, gallup. However, I think that despite this poll being a outlier, it would serve McCain very well to the extent that the media will be talking about it along the lines: "What is wrong with Barack." I don't think the media will care about the discrepancy between one gallup poll and the other. The media is loving the "horse race," and this poll fits them like a glove.

____________________

jsh1120:

I have to say that Gallup is not doing much for their credibility by including an LV screen in only one of their two surveys and then discounting its importance. Add to that the fact that Frank Newport has once again retreated to the "statistical noise" explanation for the differences in their national USA Today and their daily tracking poll, and the result is, to be diplomatic, confusing.

____________________

Undecided:

According to the linked website....As for the difference between the tracking and USA TODAY/Gallup polls, Frank says not to read too much into it. "Statistical noise" may be largely to blame.

____________________

Alan Abramowitz:

Gallup is showing a much larger difference between RVs and LVs than between their initial sample of voting age population and RVs. That makes no sense. Differences between voting age population and registered voters are always much larger than differences between registered voters and actual voters in terms of interest, knowledge, etc. There are just a lot of things about this poll that don't make sense. As in 2000, the LV screen is being applied too soon and is skewing the results.

____________________

Undecided:

Not confusing when one understands statistical concepts. "Statistical noise" is slang/shorthand for the acknowledged variation that is found within a given data sample. Such is expected in "random" sampling.

____________________

carl29:

I understand that statistical noise has to do with margin of error, although showing the same error for 3 consecutive days, the tracking from 25-27? and two different pollsters?. I really think that it has to do more with the sample size, in June the sample size of likely voters was 1,310 and now it is 791, not to mention the sample size difference among registered voters between gallup tracking and USAToday/Gallup, 2,764 vs. 900. At least online the journalists are paying attention to the discrepancy, although most of them seem to agree that this outlier must be have the McCain camp happy because at least it brings good coverage to the campaign.

____________________

carl29:

Just as I said when Newsweek and Bloomberg came with those infamous polls showing Obama up by 15%, if this is a true movement, other polls will follow. Remember that this is not the last poll, just last week we had a set of polls showing no change since last month. Today is Monday, so I assume more pollsters will come with polls in the near future confirming or denying UsaToday/Gallup findings. Again, time will tell if that is the case. To me it is an outlier, the extreme low for Obama, just like the others were showing the extreme high for Obama.

____________________

Gallup must be feeling a little heat about this poll, as they have posted an explanation of their likely voter model. You can see where the model may slip, as it probably underestimates voting among young people. The second question that comprises their three-item likely voter index is a vote frequency question, "How often would you say you vote -- always, nearly always, part of the time, or seldom?" One can already imagine the response of an 18 year old: never.

And Alan, do you have the sample size for the voting-age population? I'm not seeing those VAP results anywhere. I'm curious as to what Gallup believes will be the VAP turnout rate, which we can infer from the VAP and likely voter sample sizes.

____________________

Undecided:

The number of 18 year olds in such a polling sample is relatively small, and therefore would not have a significant effect on a likely vs. registered voter.

____________________

marctx:

Wow. The Obama channels news coverage tonight was really great for McCain. They talked the entire night about Obama's failure to acknowledge the surge and McCain's ad about Obama not visiting the troops. As a McCain supporter I agree this was a Cotto-Judah low blow, but just getting the Obama and the media trying to explain it only hurts Obama with those voters he struggles with in PA & OH. The ad was pure bull, but great politics. We need a few low blows to win this thing.

____________________

marctx:

For those of you who don't follow boxing, Cotto low blowed Judah two times and won the fight.

____________________

brambster:

LOL!!! The reliability of polling is surely suffering this season. Between the surge in partisan Republican pollsters attempting to make news of their own design, and the inability of traditional pollsters to figure out how to weight their polls for what is clearly a new twist on the electorate during a time of rapid change in voter ID, it has become pretty much impossible to take any single poll for even it's own face value.

I do trust poll averages within about a 2 point MoE where there are plenty of results (such as national polls), but beyond that it it is fruitless.

This race hasn't moved hardly an inch since two weeks after Hillary dropped out, and there really hasn't been a reason for it to move either. It's that simple. Maybe, just maybe, Obama might have made the first real move with his trip, but these polls seem to lag such events by 1 to 2 weeks before seeing the full effects.

____________________

Undecided, I think you took me too literally. I was speaking about younger people generally, not just 18 year olds, which was a motivating example. Younger people have a much lower likely voter profile using an index that is 1/3rd composed of a vote frequency question. This is likely a (major) contributing factor to the large divergence in candidate support between the Gallup's registered and likely voters. (Perhaps if we can get our hands on the actual crosstabulations for the USAToday/Gallup survey, we can know more.)

Likely voter models typically have two components. For each voter, pollsters construct a voting propensity index using a set of questions that probe interest, past voting, and other questions such as knowledge of their polling location. They then order respondents on the index and select into their likely voter universe a percentage of people equal to their predicted turnout rate.

I believe that this technique will be sorely tested this election, where we may reach a turnout rate not seen in a century of American politics. If we top the 63.8% rate of 1960, the next previous highest turnout rate would be 65.7% in 1908. Not only will predicting the turnout rate be a challenge, but the likely voter index questions may not capture vote propensities of people who will vote but do not fit a traditional likely voter profile, such as younger voters who are mobilized by Obama's campaign.

____________________

TheVoice99:

That's all the clowns have left...."low blows". Umm, good luck with that.

Anyhoo, whoever said it was "statistical noise" apparently has never taken a statistics class.

What is rather apparent is that USA Today takes trends of voters into account when anticipating "likely" voters. Too bad this election year many people who usually don't vote, will. The republicans can break out all their dirty tricks like purging voter rolls and suppressing minorities from voting (like they do every year), but it still won't do any good. Their time is over. 2006 started it, 2008 will reinforce it, and 2012 will strengthen it. Too bad marctx, your sorry chimp has come home for good.

____________________

Lechuguilla:

First it was the Newsweek poll that seemed suspicious. More recently, the Quinnipiac polls in Minnesota and Colorado seemed to contradict comparable polls. And now ... Gallup.

I get the feeling that most or even all of the major polling firms are still "tinkering" with their methodologies. For example, should factor "A" be weighted more or less than factor "B"? You would think that sixty years of scientific polling would have rendered the methodology stable. Apparently not.

All of this tinkering conveys the impression that the statisticians responsible for the formulas have little or no confidence in their own methods.

Such contradictory results, which seem to be ongoing, do not inspire poll credibility. For if we expect anything from serious pollsters it is a methodology that is, over time ... consistent.

Lech

____________________

Arnaud:

Newport from Gallup said in USA Today :"

Obama led McCain 48%-40% among registered voters in a separate Gallup tracking poll over the same three days. Newport said differences were inevitable in two polls conducted simultaneously but were within the margin of error."

Yes, 12 points gap between 2 polls from the same society at the same period is "inevitable" or "normal" and within the margin of error???

Seriously, it's ridiculous.

For me, Gallup has lost much credibility.

____________________

sherman:

I saw Morning Joe talking about this poll and pretending that it was really a significant drop for Obama. When I saw the numbers I laughed out loud! I don't recall any poll swinging so fast in one day. Even Kerry had a long, slow drop that occurred over a few months. This poll is absurd, but god forbid the media act responsibly and tell everyone how unlikely it is.

____________________

carl29:

I think that I saw one of the most hillarious thing. I was reading the article on USAToday about their "famous" poll, and looked at the chart on the left handside. Am I wrong or they have things mixed up? It seems that they gave Obama the dark blue color and McCain the light blue color. However, they either mixed up the colors and the candidates or the poll is crazier than we thought. According to the chart, Obama has the edge in dealing with Terrorism by 16%, 48% vs. 32%. Do you think so? McCain losing by such margin on Terrorism

____________________

sherman:

LOL! Yeah, the colors are switched. Terrorism is the only thing that McCain has a clear lead. What must be scary fro McCain is how close Obama has come on the Iraq question (40-42). As an aside, I've never been to Usatoday.com before. I must say their website needs work.

____________________

Undecided:

Do not confuse margin of error with variance between polling samples (aka random samples). "Newport said differences were inevitable in two polls conducted simultaneously but were within the margin of error." These inevitable differences are what he termed ''statistical noise."

Dr. Frank Newport (head of Gallup) was the person who used the term "statistical noise" and I would bet the farm that he has taken more than one statistics class.

Regarding media response... MSNBC was still reporting the Gallup tracking poll from the previous day (Obama up 9 points) and ignoring the other polls.

____________________

Undecided:

If you took the margin of error from the USA poll and looked at the range... then did the same for the Gallup daily tracking... you would discover that they overlap. In other words random samples do not neatly overlay each other.

Not also that the daily Gallup tracking poll is for "registered voters" and 48%-40% compares reasonably to the "registered" voters from this poll (Obama 47, McCain 44)... a difference of 1 point for Obama and 4 points for McCain (no 12 point differences here)... and within the margin of error.

____________________

Undecided:

I agree with marctx that the Obama campaign having to explain in detail not meeting with the wounded soldiers in Germany was not good for Obama. Too much nuance and detail involved. One GOP consultant framed it as multiple explanations and he did not know which one was the truth.

I do not agree that the McCain ad was a low blow... and the media really took McCain to task for the false statement about not bringing campaign staff/cameras... Again a framing issue because one of the early Obama campaign statements used similar words in a blanket statement (which included blaming the Pentagon) but apparently that message was not well-stated... thus the multiple followups. McCain's ad was out of the can before the nuanced explanations from the Obama campaign appeared.

Singling out McCain for low blows is hardly equitable, when in fact Obama uses a similar "low blow" (reframing the issue) in his energy ad and talking points saying that McCain did not vote for "alternative energy" in 2005 when in fact Obama is referring to McCain not voting for the Bush/Cheney energy bill. That is how Obama explains his vote FOR that bill... i.e. voting for alternative energy. Yet, he accuses the guy (McCain) who did not vote for the Bush energy policies of continuing Bush energy policies. So I say... all is fair game in politics.

____________________

Undecided:

Notable in analysis of this poll is the following statement: "The poll, taken Friday through Sunday, showed a surge since last month in likely Republican voters and suggested Obama's trip may have helped energize voters who favor McCain."

No need for voter suppression this year (Kathleen Harris and Ken Blackwell are both gone, gone, gone... Hallelujah).

The key is energizing the base... Already the GOP have "conservative values" initiatives going on the ballot in several states (apparently a Rovian scheme)... and ironically other states have "liberal value" initiatives which will also serve the same purpose... to energize the GOP base.

Historically, the GOP are consistently high in voting numbers --as well as southern blacks... which is why increasing the number of black voters in the south will be less impactful then Obama's strategy implies.

The youth vote is the key, but they failed Kerry (great early support) and did not show up. Ironically, during the Vietnam War they failed McGovern and voted for Nixon instead. Will they show up this time? Who knows? Some polls has showed their enthusiasm waning since the Democratic primary ended. And some of that enthusiasm was also damped down by Obama's move away from the left. Like any rock star, Obama needs to release a new album to keep his youth fan base. What I am saying is that youth are not very dependable and are actually very fickle... especially when it comes to showing up to vote.

____________________

jsh1120:

Undecided (said):
Notable in analysis of this poll is the following statement: "The poll, taken Friday through Sunday, showed a surge since last month in likely Republican voters and suggested Obama's trip may have helped energize voters who favor McCain."

I find that statement to be as puzzling as much of the rest of the USAToday/Gallup results, including those noted by Prof. Abramowitz regarding samples of adults, registered voters, and likely voters.

It's certainly possible for an event to have a stong net effect in one direction or the other, but I can't offhand think of a single event that had a strong net effect in one direction in one poll and a strong net effect in the other direction in another unless sampling or question wording is at fault.

Are we really supposed to beieve that Obama's trip energized his supporters in the Gallup Daily tracking poll and energized McCain's supporters in the USA Today poll?

And for that matter, are we supposed to believe that a trip where a plurality of respondents reacted favorably and only 25% reacted negatively "energized" the GOP?

Unfortunately, Frank Newport's efforts to deal with the mini-firestorm of criticism doesn't help.

____________________

Undecided:

The issue of "energizing" was specifically related to "likely voters," which is irrelevant to the Gallup daily tracking of "registered voters."

One cannot compare "likely voters" from one poll with "registered voters" from another poll... even if they both were done by Gallup. Very importantly, the results of "registered voters" differs very little between the two polls... so there is NO evidence of "screwiness."

The "mini-firestorm" is from lack of knowledge about statistics... Dr. Newport's dealing with that is in effect trying to educate those ignorant of statistics (or forgot the lessons learned from their statistics class).

Dr. Newport also stated that at this stage of the campaign that registered voters are more relevant because the "likely" voters are more changeable depending on their enthusiasm. (And right now the Republicans happened to be enthused as a reaction to Obama's trip.)

Again... "likely" is NOT the same as "registered" voters. Enthusiasm for a candidate is a measure more relevant to the phenomena that occurs if voters are likely (or not likely) to vote.

____________________

The problem with the enthusiasm explanation is that enthusiasm is not one of the three questions Gallup is using to identify likely voters. Yet, I've noticed a lot of pundits interpreting the likely voter results in this manner. Gallup's current likely voter model a three item scale consisting of (1) thinking about the campaign, (2) self-reported past voting, and (3) intention to vote. Item #2 is heavily biased against Obama young supporters, thus the divergent results between Gallup's likely voters and registered voters.

And as a little bird reminded me, Gallup has had a very high level of volatility in their likely voter model in past elections (for an analysis of the 2000 election, see Erikson, Panagopolous, and Wlezien, Public Opinion Quarterly, 2004). Expect more strange behavior out of Gallup's likely voter results in the weeks to come.

____________________

Please be patient while your comment posts - sometimes it takes a minute or two. To check your comment, please wait 60 seconds and click your browser's refresh button. Note that comments with three or more hyperlinks will be held for approval.

MAP - US, AL, AK, AZ, AR, CA, CO, CT, DE, FL, GA, HI, ID, IL, IN, IA, KS, KY, LA, ME, MD, MA, MI, MN, MS, MO, MT, NE, NV, NH, NJ, NM, NY, NC, ND, OH, OK, OR, PA, RI, SC, SD, TN, TX, UT, VT, VA, WA, WV, WI, WY, PR