Articles and Analysis


POLL: Newsweek National

Survey of 1,203 registered voters, 592 registered voters that identify or lean Democratic, interviews conducted 4/24-25 (article, results).


Among Registered Voters:
Vote Preference:
Obama 47, McCain 44
Clinton 48, McCain 45

Obama 53/40
Clinton 47/49
McCain 51/41

Among Registered/Dem-Dem Leaners:
Obama 48, Clinton 41



Can someone explain why we pay attention to national general-population polling in a presidential election? Al Gore won the last national poll in 2000, and all he got for it was a lifetime supply of Big Macs. Seriously, the major flaw with these polls is not statistical (although many of them fail to adequately disclose the difference between a Census-balanced pool and a voter-weighted pool), it's that the media reporting of the results seems intentionally designed to obscure the fact that they are irrelevant.

For example, the Newsweek article accompanying this poll states: "After an important primary win in Pennsylvania, Hillary Clinton has reduced Democratic rival Barack Obama's double-digit lead among registered Democrats and voters leaning Democratic by more than half, according to the latest NEWSWEEK poll. ... The survey found that Clinton now trails Obama by seven points, down from 19 just one week ago." Using terms like "double-digit lead" and "now trails ... by seven points" give the impression that these numbers have some bearing on a real race.

The presidential race generates a mountain of meaningful statistics and numbers; there's no need to generate numbers that don't have a real bearing on the race itself. When I see stuff like this, it merely confirms my view that the media creates this stuff for their own promotional purposes. How does this bear on the stats/polling community? It explains why some polling organizations seem to place more emphasis on generating a headline for their client than they do on producing data that is statistically reliable.


Jonathan S. Fox:

There are two major reasons that leap to mind to have national polls.

Setting aside all else, an election is fundamentally about picking the leader that the people want. The details in the way we organize the state level elections and weigh states and so on are certainly important, but ultimately the idea is rooted in choosing the more popular President. Regardless of the mechanics by which we tweak the formula, a national poll is the most direct way to measure that sentiment.

The second reason for national polls is that even if you were to dismiss the practical relevance of a national poll, it's much more expensive and time-consuming to get reasonable sample sizes in every state and break down an estimate by electoral votes or some other metric that would be considered more practical. By taking a national poll, a larger sample can be collected and a trend line built that offers an at-a-glance view of the popularity of the candidates. Headlines at this level are easier to get, but they aren't meaningless; not only do they very directly measure the basic public opinion that the presidential election is meant to respect, they're simply abstractions and a bird's eye view that do respond to and incorporate the real movement in individual states and regions.

The correlation between winning the popular vote and winning the electoral college is extremely high, and of the inaccuracies that pollsters wrestle with, there are far bigger ones -- Al Gore's recent counterexample, though fresh in memory, was a rare phenomenon. Polling on the popular vote may not be precisely imitating the mechanic by which a president is selected, but it's certainly not irrelevant to it, and a substantial and clear lead in the popular vote is almost certain to indicate an underlying lead by any more precise metric.



If you like at the unfavorable/favorable of Barack Obama you see that he seems to do far better in newsweek's poll than in Gallup or Rasmussen. This tells me that newsweek is surverying a slightly different group of peoplee. In my mind this makes me think newsweek overpolls for obama in all its surveys. BUT, maybe in this year newsweek is actually polling the right set of voters, and the other polling firms are wrong. (i would not think this is the case, but it is very possible)



National polls like this are not all that accurate or relevant this early in normal election years. This year they're especially so. Never in history have 2 leading candidates for a party nomination had such polarized demographic support. Whether or not you count the vote in FL and MI (or the primary in WA), Clinton and Obama are in virtual tie. Obama has won more states and pledged delegates, but Clinton has won more electorally rich states and more Democrat votes. She polls better in the big swing states, but he may bring a few blue/purple states "into play". What is very interesting about this new poll is that even with much higher negatives, Clinton actually does slightly better against McCain than Obama. This would indicate that many people like Obama, but don't think he's ready to be president and many people don't like Clinton, but still think she's more qualified and would vote for her anyway. This supports what I have come to believe after reviewing exit poll after poll in state after state: the main reason 15%+ of Obama supporters say they will not vote for Clinton is because they 'don't like her' (but they don't like McCain either). But the main reason 25%+ of Clinton supporters say they won't vote for Obama is because they don't think he has the experience or qualifications to be president (but McCain does). Therefore, it'll be much easier to sway the Obama supporters to Clinton (esp. if he's her running mate, which he would be) than for Clinton supporters to vote for Obama (esp. since he will not offer the VP slot to her).



Oh, Patrick, does the endless Clinton spin (lies) never tire you? Show me a national poll where Clinton does better against McCain than Obama - above the margin of error, and I'll show you a dubious poll. Yeah, that's what I thought.

You forget to mention, "electorally rich states and more Democrat votes" do not mean ANYTHING. According to the rules (yes, those things Clinton fanatics never know the meaning of), only pledged delegates matter.

I love hearing how polls now should be taken into account for November. Hmm, what did the polls say 5 months ago again?

Everything about match-up polls and "electorally rich state" polls is BS. They mean NOTHING. Talk to me in August/October about polls. Super delegates aren't stupid (not all of them anyway), and they will not take any of those nonsense polls into account - that is why they will back Obama - the winner of the most pledged delegates.

Now, if the bogus claim is that super delegates should overturn the will of the people (Obama has won double the number of primaries and caucuses and also will have the pledged delegate lead), based on polling and "electability" then WHY HAVE ELECTIONS AT ALL?? In order to be "electable", you need to win ELECTIONS - something Clinton has been unable to do against Obama. Why would she fare any better against McCain??

One thing you conveniently forgot about polls the past few months:

Obama ALWAYS does better than what the polls initially say - after campaigning in a state extensively.

Has Clinton EVER done this? Look at PA, she had a 20 point lead that evaporated to 9 in a matter of weeks.

That is why polls now will not be ANY kind of determining factor for the super delegates. If anything, they would factor in the Obama-effect and see that he would come out with over 300 electoral votes come November (look at how many states are in-play already) - and by that consequence add substantially more Congressional seats for the DEMS than Clinton would ever be able to - even in her wildest dreams.




Jonathan's comment ("election is fundamentally about picking the leader that the people want") is true for all elections EXCEPT the POTUS. Historically, the process of picking the POTUS was specifically designed NOT to be about respecting "basic public opinion," but rather about the States (ie. as political entities, not as collected citizens) choosing the best executive. Hence, citizens don't vote for the POTUS; states do, and they don't vote in direct proportion to population, and then they don't even vote directly - they send electors. The country has pushed the process slightly towards populism by the custom of selecting electors through popular voting, but that's not required (any given state could choose electors differently if it so desired). Given the recent experience of not-quite-President Gore, I assumed everyone took this for granted. I don't view Gore as a counter-example; but rather a very real reminder that national popular opinion is not causal to winning the POTUS, it's correlative in a second degree sense.

Jonathan's second point is the one I was really interested in - why don't we focus on state-by-state polling in the key states? Why do we choose to include Vermont and Idaho results in the national sample? (Anyone want to give me odds on those states going against their 30 year trend?). Having a head-to-head poll of Obama versus McCain in 12 key states would be much more interesting from an election standpoint than knowing the national popularity contest result.

I completely agree that the national poll is a much easier way to get an "at-a-glance view of the popularity of the candidates," and my point is that this tabloid-level information is what the media want, and not a more serious measure of the election.



I meant to post the following here, but in the process of registering to be able to post I was taken to the next day's polls. Sorry for the "recycle."

I think that those who feel that pointing out Obama's weakening poll numbers is not "helpful" should go to a place where they would get what they wish to hear. This site provides info on whatever is out there and all indications are that Obama has lost his "shine".

The superdelegates must take this into account, otherwise there would be no rationale for the whole concept of superdels, if all they do is to ratify the status quo. Obama would not reach the number of delegates needed to win the nomination without the superdels, which means that their existence is not predicated upon endorsing whoever is "ahead" in pledged delegates. Their "raison d'etre" makes sense only if they would help choose someone who has the best chance to lead the party to a victory. At this point, Obama is looking less and less like that someone...according to the data posted here, which is why his supporters are saying that it is not "helpful" to point that out.


Post a comment

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.