Since I’m for McCain, I was pleased to read the Arizona Senator has evened things up with Obama in the new Gallup Poll. But is this real? Only a few days ago, the LATimes was showing an Obama slaughter. What changed? Is it the pump prices and Obama’s non-response to drilling? McCain seems to have a more coherent and multi-faceted energy plan while Obama seems mired in very old-fashioned anti-nuke liberalism, but still… The question is the polls. How do we know if and when to believe them? We are looking for someone to write dispassionately about this on Pajamas Media. Feel free to leave suggestions here.
“June 25, 2008
Don’t follow the bouncing poll
Two recent polls have Barack Obama leading John McCain by more than 10 percentage points. A Newsweek poll has the race at 51-36 and a Bloomberg/LA Times poll has it at 49-37. Thanks to these polls, the RCP average now tilts decisively towards Obama by a margin of 48.2 to 41.3. This would suggest that Obama has obtained a substantial “bounce” recently.
But the polls I pay closest attention to — Rasmussen and Gallup — indicate otherwise. The latest Gallup tracking poll has the race dead even (45 to 45). Moreover, Obama’s lead in this poll has not exceeded 3 points in the past week. The latest Rasmussen tracking poll has Obama ahead by only 4 points (49 to 45 with “leaners” included). Rasmussen’s polling shows a stable race, with Obama consistently ahead by 4 to 6 points (with leaners included).
For what it’s worth, I’m with Rasmussen. There’s no reason why the race should be unstable at this point — no event of any consequence has occurred since Hillary Clinton dropped out and endorsed Obama in early June, and it’s unlikely that voters are paying much attention to the race these days. Even I’m having trouble doing so.”
http://www.powerlineblog.com/archives2/2008/06/020839.php
Powerline
June 25, 2008
Don’t follow the bouncing poll
Two recent polls have Barack Obama leading John McCain by more than 10 percentage points. A Newsweek poll has the race at 51-36 and a Bloomberg/LA Times poll has it at 49-37. Thanks to these polls, the RCP average now tilts decisively towards Obama by a margin of 48.2 to 41.3. This would suggest that Obama has obtained a substantial “bounce” recently.
But the polls I pay closest attention to — Rasmussen and Gallup — indicate otherwise. The latest Gallup tracking poll has the race dead even (45 to 45). Moreover, Obama’s lead in this poll has not exceeded 3 points in the past week. The latest Rasmussen tracking poll has Obama ahead by only 4 points (49 to 45 with “leaners” included). Rasmussen’s polling shows a stable race, with Obama consistently ahead by 4 to 6 points (with leaners included).
For what it’s worth, I’m with Rasmussen. There’s no reason why the race should be unstable at this point — no event of any consequence has occurred since Hillary Clinton dropped out and endorsed Obama in early June, and it’s unlikely that voters are paying much attention to the race these days. Even I’m having trouble doing so.
Sorry for the double post. You might want to also ask Karl over at PW.
June 18, 2008
A horserace poll quickie [Karl]
With the usual caveat about not caring much about national horserace polls, especially this far away from an election, there has been a spate of them now that are worth a comment or two:
T[ues]day’s ABC News/Washington Post survey of adults, which was taken 6/12-15, shows Obama leading McCain 48-42% — down from Obama’s 7% lead one month ago. Meanwhile, a new Cook Political Report/RT Strategies poll of RVs, also taken 6/12-15, shows Obama leading McCain 44-40%. Similarly, the latest Gallup tracking poll, which queried RVs between 6/13-15, shows Obama with a 4% lead, 46-42%.
For NARAL’s latest battleground survey, Greenberg Quinlan Rosner (D) surveyed 1,000 RVs in 12 swing states: CO, FL, IA, MI, MO, NV, NH, NM, OH, PA, VA and WI. While Obama shows leads in national polls, he fares worse in these key battlegrounds. Obama bests McCain by just 2% overall, 47-45% with margin of error +/- 3%, and ties him 43-43% among Ind women.
The first point – so obvious that it is easy to overlook — is the difference in samples. Obama’s lead is largest among adults, but shrinks in polls of registered voters and becomes insignificant among likely voters. Indeed, the ABC News/WaPo poll itself demonstrates the point:
http://proteinwisdom.com/?p=12534
All polls today need to be looked at not just for the surface numbers, but for the sampling numbers and for the question phraseology. A pollster this far out from the election can rig the sample any way they want, and there’s no way to check their accuracy — maybe Barack Obama really did lead John McCain by 15 points on the third week of June, but there’s no way to tell, or to hold the pollster accountable for their results, when the votes won’t be cast until November.
So if Newsweek wants a poll that samples Democrats at a 1.7:1 ratio to Republicans (37 /22 percent) the way their recent poll did, it really doesn’t matter — the only poll that does is the one right before the actual election, where if your polling firm calls a 15-point Obama win and McCain wins by three, you’re a laughingstock and you lose clients, because you’re also either mendacious or incompetent.
My guess is when Newsweek does their final pre-election poll Halloween weekend, they aren’t going to be using a 1.7:1 sample anymore.
In addition to the point about registered voters versus likely voters, you must also take into account the size of the sample. A poll of 800 likely voters would probably be more accurate than a sample of 1,000 registered voters. The next question is how the polling sample was weighted by party. I believe that the current weighting for the US is Democrats 37%, Republicans 32%, and Independents 20%. The balance are the remaining parties that are out there. If a polling sample includes 41% Democrats, 25% Republicans, and 34% Independents, it will give an inaccurate result.
Additionally, another factor could be how the questions on the poll are worded. This could push voters towards one candidate. There is one other factor to consider: a large number of democratic voters have stated that they will stay home/vote for McCain/write in their candidate if their candidate doesn’t get the nomination. I believe that the figure I saw was that 28% of Clinton supporters and 20% of Obama supporters would go the other way.
Lastly, it’s the summer. Things are notoriously uneven, and people aren’t focused on the election just yet. Let’s see how things look in September before the accuracy of the current polls can be judged.
Polls ARE gibberish. They “interview,” with their oft-biased questions, about 920 folks over two nights. I have never been called, have you? Has Joe in Dubuque? Or is is mostly Jamal in Brooklyn? I’m just wondering.
Yes, McCain has a better plan—everywhere. Obama is a naive, radical simpleton, to put it bluntly. Dangerous to the USA indeed.
But the only polls that matter are conducted every four years, and in seven of the past ten, the pro-American party (the GOP) has triumphed.
A lot of urban, coastal folks need to realize that America is not made up on liberals and elites. It is made up of patriotic folks who support our troops and loathe our enemies. They also see thru Obama. Maybe the yuppies at Starbucks don’t, but most do.
John Allen Paulos writes well about innumeracy. I am not sure if he has examined the problem with polls, but no doubt he could provide a thoughtful and entertaining article.
I thank dre for recommending me. The Gallup Tracking Poll was close even during the periods covered by Newsweek and the LAT. The difference appears to be in the sampling. Though most polls show a party ID shift toward the Dems, the samples for the outliers are skewed even more than that. Plus in one (the LAT, iirc) the sample skews quite young.
This actually points to a problem pollsters have had this cycle and will have thru Nov. In estimating likely voters, how will the account for young voters (esp. first timers) and the “enthusiasm gap” (if it lasts to Nov.)? Usually, likely voter screens downplay younger voters (as they are largely based on past voting) — and there is no real way of knowing if that underestimates Obama’s strength this time.
Hope that’s helpful.
1948 – “Dewey Beats Truman”
2004 – “Kerry Beats Bush” – exit polls no less.
And yes, I do remember the 1948 headlines.
Ninety eight percent of the news is either polls or pretty blonds in trouble.
I think I’ll wait for the real poll.
I have never been galloped
They’ve probably attempted to poll me, but I hang up after hearing the machine click on and before the person starts talking.
I don’t have a land line phone. Neither do a growing number of people, especially the young–polls mean less and less with each passing year.
The Gallup is a tracking poll. It has been consistent and moved within a much shorter range. The newsweek poll is garbage.
I think that one must take into account the possibility that the definition of “likely voter” may not be accurate for this election. Obama has energized certain categories of voters who may vote at much higher percentages than previous elections.
That LA Times poll had Democrats outnumbering Republicans about two to one. That is just ridiculous.
And even then Obama could not get to 50%. Strange.
Newsweek was much the same. They just make sure there is a demographic advantage. It makes all the difference.
I think Obama might well be ahead right now, but not so much.
We have to remember the purpose of the polls: to make people think that they will be in a minority if they don’t vote for the favoriate. After weeks and months of publishing that Obama is ahead in the polls, they hope they many people will vote for him just to go along with the majority. Most people don’t like to be a minority. But let’s hope the people they are trying to influence either are immune or won’t vote anyway.
I was recently called and asked this “unbiased” questions: Do you think the police should be arresting people for medicinal marijuana use instead of going after dangerous criminals? I said yes, and the pollster sounded very disappointed when saying thank you and hung up.
But how do you answer a question like that? Instead of? Of course not, but if I said no they would have recorded it as a yes for medicinal marijuana, so hopefully beat them at their own game.