Articles and Analysis


Chuck Todd's Challenge

Topics: 2008 , Barack Obama , Bob Barr , Clinton , CNN , John McCain , NBC News , Pew Research Center , Ralph Nader

For statistical modelers, number crunchers and turnout scholars among our readers -- and you know who you are -- Chuck Todd and the crew at the NBC News political unit have thrown down the gauntlet:

*** Analyzing the turnout: After crunching numbers for the last several months during the Clinton-Obama contest, we’ve been experiencing mathematical withdrawals now that the Dem race is over. In a word, we have the shakes. So to calm our nerves, we got out our abacuses and did some initial fooling around with projected popular vote. Using the 2004 results as a baseline, we were curious as to which states would swing to Obama if he does raise overall turnout by 20% (approximately another 22 million voters) and wins those new voters by a 60%-40% split. Assuming an even distribution -- which we know is potentially a flaw in this estimate, so back off! -- a 20% turnout increase breaking 60%-40% for Obama would swing four states from red to blue (Iowa, New Mexico, Nevada, and Ohio). If Obama wins the new voters by a 65%-35% margin, two more states come over (Colorado and Florida), with another (Virginia) essentially too close too call. We're going to crunch these numbers a number of ways over the next few weeks, including using the 2000 election as our baseline (since many folks believe 2004 over-estimates the GOP electorate); seeing what would happen if Obama runs a 50-state campaign but McCain runs a 17-state one; and finding out what the realistic maximum population vote advantage Obama could have while losing the electoral college. In the meantime, have fun with this model.

Thoughts anyone? Alternative models or analysis? Post a comment or email it our way (questions at pollster dot com).



I would first refer you to 538 - they found that McCain, not Obama, is more likely to win the popular vote but lose the election, because the fastest growing states are mostly in his column, and are underrepresented because the electoral college is based on the 2000 census.

NBC's experiment is not that surprising (coming from them) it is essentially "what if Obama wins!?" You need to figure out demographic groups - for instance, does Obama do worse among old people or women?



The major flaw in this "analysis" is that it would require in increase in votes to change a state like Colorado to the Dems when Obama is already up by 10 points there.



Mark Lindeman:

I don't know whether Todd noticed another crucial assumption of the model, which is that everyone else votes (roughly and proportionally) the same way in 2008 as in 2004. In 2004, Kerry won the 2000 non-voters (not 60-40, to be sure) but still did worse in the popular vote than Gore did in 2000.

By the way, we don't know Kerry's margin among these voters who did not vote in 2000. The exit poll says 54-45, but it's clear that some people misreported their past vote (or non-vote). Most likely the actual gap was smaller.

In 2008 McCain doesn't have the incumbency advantage (although he does have a deeper reserve of good will than Obama), so I don't necessarily expect him to gain ground among returning voters. I don't expect him to perform identically among them, either. I don't think it's ever safe to focus on one kind of voter as the key to victory.



Breaking News!!!! (Sorry for those leaking their chops with Sen. Obama's defeat).

GALLUP: Obama Gains Among Women After Clinton Exit. Obama's lead among women has now expanded from five percentage points to 13, Obama 51%, McCain 38%. When Hillary was in the race, she led McCain by 52% to 40%.

Obama has also seen his support among non-Hispanic white women increase modestly, though McCain still holds a slim 46% to 43% advantage among this group. Prior to Clinton's departure, McCain led Obama by 50% to 41% among white women.

Obama's current 13-point advantage over McCain is essentially the same advantage that Clinton held over McCain throughout her active candidacy.

Hispanics? checked!
Women? checked!

In one week!!!!! Go Obama

P.S: Interestingly, while white women were a major part of Clinton's primary constituency. She only tied McCain among white women, at 47%. White women right now? McCain 46%, Obama 43%. Hasn't the healing started? Any thoughts?



Carl, have you been eating pixie sticks for breakfast again? Obama went up 2% among white women from 41 to 43, while McCain fell a bit. There is a reason the article says "modestly". There is also an increase in undecided white women, from the sounds of it.



Oh my dear eternaltriangle: McCain has seen a drop of 4% of white women support and an Obama increase of 2%, there is a 6% swing. Actually, against Hillary, a white woman, white women supported McCAin by 47% and now against Obama, a black man, white women are supporting McCain 46%. He is losing them quite fast. Keep in mind that Hillary left a week ago. A 6% swing in a week is as fast as it can get.

My theory has always being that these women will move gradually. First phase, they'll say that they are going to vote for McCain out of anger. Remember that when Hillary was on the race, McCain gathered 50% of white women support against Obama. Second, as things cool off, they will move to undecided. Now, McCain is supported by 46% of white women. Third phase, white women will come move over from undecided to the Obama camp, once Hillary starts campaigning with Barack.

To be honest, this has been faster than what I'd thougth. It has been a week for goodness sake!!!!

P.S: Welcome Girls!!!



it doesn't matter eternaltring

this puppy is over.....

cottage cheese in lime green jello has no shot.

if I were the repubs, I would just stop now and save some money...whats the point?



Another fact which I think is extremely important:

The hardest group will be women over 50.

As to June 2 among women over 50:
McCain 46%, Obama 43%
Clinton 51%, McCain 41%

Now, as to June 9:
Obama 47%, McCain 41%

In a week we can see a 9% swing. 5% of women over 50 have dropped McCain and 6% of women over 50 have picked up Obama. Actually, as to now McCain level of support among women over 50 is the same level of support he had against Hillary among this group. I think that people are calming down very, very fast.


I suppose that the question is framed as "what will it take for Obama to win?" since the baseline comparisons are the 2000 and 2004 elections.

I think Mark Lindeman's comments are on target, and imagine that Chuck Todd's spreadsheet has multiple moving parts. Would he be willing to share it with the pollster.com world so that we can dissect it? Perhaps Mark Blumenthal or someone else would post an approximation of what Todd is doing, if he is concerned that someone would tag the sheet an "NBC's official projection"?

I'll have to think on this, but my initial addition to Mark Lindeman's comment is that non-voters come in many stripes. The people most likely to be activated in 2008 are the occassional voters - people who have a past history of voting but do not turn out for every election. These people may have a different candidate support profile than people who rarely vote, but are (or have been) registered, and both may be different than those who are not registered and have never voted but may be activated by their campaign interest or campaign mobilization efforts. (I wish that I could encapsulate this into a nice soundbite, but this way of categorizing non-voters helps explain why Bush was successful in Ohio despite high turnout in 2004.)

I thus imagine that Obama's overall support among "new voters" not present in 2004 changes depending on who is activated. My suspicion is that because Obama has support of his base, he will be more able to run an effective ground campaign to turn out people other than occassional voters. McCain needs the support of his base, not just to vote for him, but also to help him organize a ground campaign. I haven't seen the same levels of enthusiasm among McCain's supporters as I have among Obama's, so this is something to monitor as the November election approaches. There is still plenty of time for McCain to bring his campaign around since there is ample evidence that most new voters register near their state's registration closing date.


Alternative models or analysis?

I have no data in hand to offer, but there seems to be a massive "intensity" gap between backers of the two candidates. If you split the voters and look at Democrats, I think it's possible to make some assumptions about what they'll do based on 2000 and 2004. In 2000, they weren't particulary motivated and the Democratic coalition remained frayed. In '04, they were pretty highly motivated, and Kerry turned out a massive electorate. (Just not quite as massive as Bush.) Will Obama turn out all those Dems again in '08? Will he add independents and swing voters (that is, dip into the Bush's pool). What effect will local elections, now overwhelmingly swinging Dem, have on voter turnout?

All of that is pretty straightforward. But what do you make of the GOP? Their candidate is radically different than Bush. He re-jiggers the GOP coalition in ways that make it seem unwise to use '04 numbers as a guide. Will he hold evangelicals? Will they put together a grassroots effort on Get Out the Vote campaigns? What percentage of those highly motivated '04 voters remain highly motivated? What percentage will turn out?

My prediction is that if the election is close, it will be because a record number of voters turned out, and both candidates were appealing to them. If GOP voters don't turn out in what has the makings of a very grim election, it won't matter how many new voters turn out on the Dem side--the overall percentage of voters won't increase much or at all. And in that scenario, Obama will attract a bunch of new voters, McCain will lose some, and it will be a blowout.



This is pretty much the question that Nate Silver addressed a month ago in outlining a strategy for Obama to win based especially on increased turnout among African Americans, Latinos, and young people.

See "Black, Youth and Latino Turnout, and Obama's Electoral Map," FiveThirtyEight.com (MAY 11 2008):



A further thought...

Are there polls out there with likely voter screens being run for the general election? Now, I'm not a big fan of likely voter models this far out in front of the election, but examining candidate support levels as the turnout rate increases would be a way to get some purchase on Todd's question.



Michael, from your website (http://elections.gmu.edu/) one might conclude that the much-hyped huge jump in turnout in the 2008 primaries is something of a myth. I believe you've published on this as well with respect to SuperTuesday. Have you done any direct comparison of turnout 2008 vs. 2004 and 2000 by state, adjusting for other factors such as whether there was a simultaneous primary in both the Republican and Democratic parties (with "viable" candidates)?


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.