2012 Forecasting: Obama v. Palin

Interesting! And what do you think about Sarah Palin's pic on Newsweek cover? (0.00 / 0)

Sometime I wonder if a monkey would be able to pick a winner better than this kind of methodology.

Regarding Sarah Palin, she pisses me off pretty much every time she opens her mouth. But for once, I thought she was right when she decried the Newsweek magazine cover picture. I think it was totally uncalled for that the magazine put that picture. And I think it was kind of sexist. I was hoping that DCW would have taken a stand and shown that despite disagreeing with her on most issues, she is right on this one.

---

by: nouvelle_ange @ Wed Nov 18, 2009 at 16:26:07 PM CST

I disagree, here's why (0.00 / 0)

I do NOT think that the cover is sexist. Palin is NO LONGER a government official - she is an ordinary citizen so there is no reason that photo should not have been used.
Also of note it was TWO WOMEN that picked the photo to be used on the cover.
I believe Palin is hollering sexism 'only' AS A DISTRACTION to take the focus off of the fact that she is ABUSING and DISRESPECTING the U.S. Flag in that photo on the cover of Newsweek!
Just my opinion  ;)

---

by: Leah @ Wed Nov 18, 2009 at 17:06:59 PM CST

[ Parent ]

The problem isn't that it's sexist (0.00 / 0)

I think the problem with the cover, if there is one, is that it's frivolous. Palin, whether you like her or not, has a large amount of support among the Republican base. She's a former Governor, and she should be treated as a significant politician in this country. I think the cover wasn't fair to her - it didn't treat her seriously. She should be treated seriously, so she can then fall flat on her face.

---

by: Matt @ Wed Nov 18, 2009 at 19:49:28 PM CST

[ Parent ]

but she doesn't deserve to be treated seriously! (0.00 / 0)

---

by: LindaS @ Fri Nov 20, 2009 at 19:17:20 PM CST

[ Parent ]

Scoff if you will (I was skeptical at first, too.), but... (0.00 / 0)

...this model performs as well or better than some other presidential election forecasting models.  If this were a situation as in the literature on the impact past political experience has on congressional electoral success -- where there have been similar efforts to, for lack of a better term, over-quantify the concept of experience -- then perhaps there would be something to you complaint.  When experience is over-categorized it doesn't do as well as a simple binary variable indicating whether a past political office was held or not.

But that isn't the case in this instance. This model outperforms the more simplistic models based on economic indicators and polling.  

This monkey says just read the original paper.

---

by: Josh Putnam @ Wed Nov 18, 2009 at 19:20:46 PM CST

[ Parent ]

BUT (0.00 / 0)

I believe the model does not take into consideration many factors which might sway a person's vote in a presidential election in our modern age:

1) A female candidate for the presidency (not VP)
2) Cable TV and Talk Radio
3) New media: Facebook, Twitter, Blogs
4) The modern 'branding' of a candidate

We really don't have ANY past elections that include ALL of these factors in a presidential race, so it is still premature to say that this forecasting model will perform well in future presidential campaigns.

---

by: Leah @ Wed Nov 18, 2009 at 19:55:12 PM CST

[ Parent ]

I'm conflicted on this to some extent. (0.00 / 0)

By my nature, I'm a big fan of parsimonious research; simple, elegant, explanatory.  And while I won't make the argument that this forecasting model somehow treads the balance between that and an everything but the kitchen sink approach (it trends toward the latter), I do think they've got enough included here to more than adequately cover their bases.

Does it cover some of the modern era factors you mentioned? No, but you can't tell me that there's a whole lot in common between McKinley's election in 1900 (the first election on which they tried the model) and Obama's election a year ago.  Things changed across that period; drastically so, I'd argue.  Yet, the model as currently configured only missed on three elections during that time.  

In the end, their objective wasn't to throw everything in that would affect a person's vote choice.  They were simply taking biographical information on the candidates as a means of forecasting presidential elections.  And it works pretty well.

---

by: Josh Putnam @ Wed Nov 18, 2009 at 20:34:34 PM CST

[ Parent ]

Well... (0.00 / 0)

The model might hold up when a race is between TWO MEN.

BUT
I don't think that the model will work regarding a race between someone like Obama and a female opponent.

A race between a male and female:  Height and weight will not matter (which is included in the list).  And in the case of Palin her level of intelligence and knowledge is not a factor with her supporters hmm cult members.

It's different when you're just comparing two men - like back when Nixon had to go up against the good looking tan tall JFK.

I would wager to bet that Obama would win if up against Palin but it would not be because of anything that is on list, he'd win because most of us know that it would be the end of the world as we know it if Palin were to win!   LOL  :)

---

by: Leah @ Wed Nov 18, 2009 at 20:52:38 PM CST

[ Parent ]

I agree, Leah. (0.00 / 0)

There are some potentially gender-specific factors included on that list that may lead to some problems for the model overall.  But that seems to me to be a pretty easy fix: just remove those factors.  One of the drawbacks of this particular piece is that it isn't ever really explained why they settled on those 49 factors.  Why 49 and 39 or 29 or whatever?  What do we gain by adding in a 49th factor that we wouldn't get with 48? The list of factors ends up being like the health care bill to the Republican Party right now.  It is long enough that you can cherry pick a handful of things that don't quite fit and draw on the negatives there despite the positives drawn from the majority of what's included.

In the end, your last comment is the key.  This is an imperfect test simply because we know the outcome already.  Obama would win.  In addition, we likely wouldn't learn much about female candidates from Palin for use in future predictions anyway simply because she would be a flawed nominee for the Republican Party.  

Is this useless? No, it isn't.  We can learn anything if we don't compare what we've got to the past.  We know for instance, that Palin's deficit on this measure is twice that of other trailing candidates who have won.  Would that change if we were to add in variables to account for gender differences?  Perhaps and perhaps very likely, but there is some value to that baseline comparison.  We won't know how much of an impact those gender-specific factors (when and if they would be added in future iterations of this model) have without it.

---

by: Josh Putnam @ Thu Nov 19, 2009 at 07:56:32 AM CST

[ Parent ]

"Gender-specific factors" (0.00 / 0)

I agree with Leah on this. It's very possible that there are more than a few factors on this list that play differently for a woman than for a man. For example, having children is a plus for a man--it gives an impression of humanity. But I bet it's a negative for a woman--people wonder if she can be a mother and a President. Unfair, yes, but I suspect it's true.

So here's my list of factors that might be significant for males but not (or even reversed) for females: adopted children, children, not divorced, loss of children, married, fraternity (presumably sorority), actor, height, weight.

That's a big chunk of the model. And since some of those factors might actually reverse for women, it's not really good enough to just say we'll leave them out and recalculate. We'd have to guess how they change for a woman, and then it's not a model with any track record at all, it's just a scorecard we've made up.

---

by: SarahLawrence Scott @ Thu Nov 19, 2009 at 08:37:18 AM CST

[ Parent ]

Scott, you're right... (0.00 / 0)

...but I'm not suggesting that you drop the problematic biographical cues for just one election or only the elections that have female candidates.  You're right: that's just a scorecard in that case. It has no comparison value whatsoever. You'd have to apply that "drop" to all the other elections as well.

Look, as I said in drawing the analogy to the experience/electoral success issue above, I'm all for an "as low as you can go" approach; the simpler the better. [Sometimes the simpler operationalization is just superior in terms of explanatory value.]  I still don't know why there are 49 cues (actually 62 in an updated version of the model), for instance.  What can be taken out -- whether gender-specific or otherwise -- and applied across the board in all elections?  I'd rather they boil this down to its essence -- to which biographical factors genuinely matter -- instead of increasing the list.

---

by: Josh Putnam @ Thu Nov 19, 2009 at 20:14:05 PM CST

[ Parent ]

Newsweek cover (0.00 / 0)

There have been plenty of photos of Obama in shorts, even shirtless. What is the problem with the photo of Palin? I do not see the photo as being anti-Palin.

---

by: loeb @ Wed Nov 18, 2009 at 16:54:06 PM CST

i wouldnt mind a shirtless picture of sarah... ok, now that was sexist:) (0.00 / 0)

i thought the picture would have been unacceptable, except it was one she had taken willingly for another photo shoot. if it is good enough for one cover, why not the other?

---

by: UplandPoet @ Wed Nov 18, 2009 at 18:22:09 PM CST

I prefer this photo (0.00 / 0)

I prefer this photo over the silly one of Palin in shorts ;)

---

by: Leah @ Wed Nov 18, 2009 at 18:28:47 PM CST

[ Parent ]

Or this one (0.00 / 0)

---

by: Leah @ Wed Nov 18, 2009 at 18:31:30 PM CST

[ Parent ]

okay, now you are being sexist:) (0.00 / 0)

---

by: UplandPoet @ Wed Nov 18, 2009 at 19:58:25 PM CST

[ Parent ]

Getting back on topic LOL (0.00 / 0)

Very interesting.  But in reference to the hoopla with the FAR RELIGIOUS RIGHT I just wonder how this will affect this model.  I am referencing Leah's post:

http://www.demconwatchblog.com...

which is a factor that I do not believe has been relative till now.  These nuts are really out there.
With this factor in place, Skin color is also not addressed in your biographical data.
Frigging 2009 and now this stuff is stronger than I can remember.  

Leah, I hope it is OK for me to reference you.

---

by: jean @ Wed Nov 18, 2009 at 18:33:10 PM CST

DocJess, are you okay? (0.00 / 0)

DocJess, I was just checking DCW for your comments on the bill that Harry Reid unveiled tonight. Can't wait to hear your point of view.

---

by: nouvelle_ange @ Wed Nov 18, 2009 at 23:57:26 PM CST

Ditto (0.00 / 0)

I must say I am quite concerned. I am not exactly sure what you were referring to in an earlier post about being really down. I figured it had something to do with the state of health care debate or the stupak amendment.

However, your lack of a post on the Unveiling of the Senate bill is really concerning. I know you as about as much as you can know someone over the internet, and I just can't believe you wouldn't be all over it unless something were really wrong.

You should know that your political blogging, and all those of the DCW staff, are greatly appreciated. I know I have said this before but I can't undervalue the perspective you offer on elective politics. YOU are appreciated.

Thanks for all you do! I hope everything is ok!

---

by: ChadCo @ Thu Nov 19, 2009 at 00:27:03 AM CST

[ Parent ]

we all love the doc! (0.00 / 0)

glad to see you are posting:)

---

by: UplandPoet @ Thu Nov 19, 2009 at 04:35:58 AM CST

Where I've Been (0.00 / 0)

I WILL get something up on the Reid bill.

I'M fine -- my dog is quite ill as I posted about a couple days ago (? - I've sort of lost track of time) -- If we're VERY VERY lucky, she gets more time. I don't know if what we're trying will work but we keep trying.

So - YES I understand my Reid obligation, but Olivia is acute, and with acute things the patient either recovers or dies quickly...I hope y'all understand.

And yes, our site was down for a few hours last night due to something with the servers.  

---

by: DocJess @ Thu Nov 19, 2009 at 05:04:46 AM CST

[ Parent ]

This seems slightly random (0.00 / 0)

The 'this works pretty well for past elections' doesn't really hold much sway with me.  First of all it isn't particularly surprising that a model would be good at predicting data that was used to construct the model.  What I mean by this is if it was determined what made a good candidate by looking at previous winners then in the end you should inevitably end up with winning characteristics similar to those winners.  Second, even if they weren't actively fudging their model in this method, it would only be reported if it ended up being fairly good.  If a thousand people tried this kind of analysis, and each tweaked their model 100 times, it would be surprising if one of them didn't end up with this good of a fit.

As a more specific question, how does Obama beat Palin 1-0 on religion?  Obama belonged to a church that was viewed as somewhat troublesome, and his defense was he didn't go to church very often and eventually had to disavow the church.  This is even ignoring the small minority that don't believe his stated religion.  Of course Palin's church is a little nutty as well, but 0-0 seems more reasonable.

---

by: pretzalz @ Thu Nov 19, 2009 at 17:23:09 PM CST

Also (0.00 / 0)

The first/last name popularity (if the name is common or not) would NOT come into play when you have two widely known names such as Barack Obama and Sarah Palin.  Perhaps if you have a local race and have two people that are UNKNOWN someone might pick a name that they thinks sounds better but NOT in a presidential race!  LOL

Perhaps the model might have worked for past president picks but I doubt it will work in the future with the addition of mass media and female candidates - the game has changed and so must the model - if there even could be a model since so many OTHER factors that can influence a race that are more important than what is on the list.

---

by: Leah @ Thu Nov 19, 2009 at 17:59:28 PM CST

[ Parent ]

Perhaps... (0.00 / 0)

...but aren't there people who probably voted against Obama or for McCain because of Obama's name.  Now, name may very well have been a proxy for something else (race, xenophobia, etc.), and that would have been problematic in a regression model if two similar factors were included.  The resulting multicollinearity would have biased the overall model.  But this isn't a regression model.

This is something far simpler than that, yet it is effective.  I still don't see what impact new mass media effects are going to have.  It can't change someone's biography.  Gender, I'll agree with you on. That could be problematic for this model and is probably why they have included it in a more recent version. [Palin would have lost a point to Obama on that front.]

In the end, a true test of this model isn't Obama v. Palin. As nouvelle_ange said at the top, a monkey could probably choose that one correctly. It isn't a difficult race to call.  The true test of any election forecasting model is how well it does in close elections.  Statistically-based models sometimes have a tough time under those circumstances. But this model correctly predicted all the close presidential elections with the exception, perhaps, of Carter v. Ford. I still prefer the statistical models -- Hey, that's what I do. -- but you can't argue with the results.

I'm just kind of sad that some of you didn't pick up on the fact that Gore was the correctly predicted winner in 2000 -- that the model is looking at popular vote winners. Oh well.

---

by: Josh Putnam @ Thu Nov 19, 2009 at 21:25:49 PM CST

[ Parent ]

Actively fudging!?! Seriously? (0.00 / 0)

I suppose it happens, but these guys are academics not partisan pollsters hired to conduct a poll for the campaign of one of their own.

What I mean by this is if it was determined what made a good candidate by looking at previous winners then in the end you should inevitably end up with winning characteristics similar to those winners.

But isn't that what forecasting is all about? Take the economy. Typically, we'd take indicators of a good economy to tell us when to look for the economy to improve when it's bad, right?  I suppose we could guess, but that kind of detracts from what we try to do: to learn. You talk about tweaking like it's a bad thing. I'm not suggesting you throw in variables willy nilly like shoe size in a model for vote choice, say.  You still have to have some theoretical justification for including any variable.

The idea is not to keep throwing things at a model to make it work.  The idea is that you start of with a theory, a hypothesis, and always come back to that in everything you do in a project.  I suspect that they started out with a basic model, took it to conferences and were asked by their peers if they had accounted for factor X or why they didn't include factor Y and it grew incrementally from there.  Google Scholar has the tag line "Standing on the shoulders of giant" for a reason: because of the power of the incremental increase of knowledge.

Obama's religion was scored one because the Church of Christ is considered a mainline religion (more members overall) than Palin's non-denominational Christian church (see the appendix of the paper for the full explanation).  

---

by: Josh Putnam @ Thu Nov 19, 2009 at 20:53:33 PM CST

[ Parent ]