This Is a Silly Question (0.00 / 0)

What exactly are you predicting? You say it's your business never to be wrong, but if today's prediction of Obama/Huckabee turns out to be right, then your previous predictions of Clinton/Huckabee or Clinton/Romney are wrong. Which one counts?

It sounds like I'm being snarky, and I don't mean to be. I just think I'm misunderstanding how your analysis framework is supposed to be used. It seems like more of a "here's what would happen if the election started today" method and less of a "here's what will actually happen" method. I think the former's very useful, but it still leaves lots of room for someone to be disappointed down the road because you can't predict what might shift the status quo between now and then.

Or is the prediction contained in the model you're using? You're predicting that whoever wins Iowa is going to have a lot of momentum in the following primaries, which makes sense to me but isn't necessarily a given. Given the closeness of the polls and the wackiness of the caucus, I can quite easily see you getting the Iowa winner wrong but still pegging the overall dynamic.

Which, as you've pointed out before, is awfully depressing to anyone who's not in Iowa.

---

by: Dave Thomer @ Thu Dec 20, 2007 at 06:04:28 AM UTC

(The sketch of) a real argument (4.00 / 2)

The argument is that you should treat today's poll averages as indicators of a set of probabilities. If you do that you might then conclude that even if Obama has the highest probability of winning Iowa, and even if an Obama win would then radically increase the probability of his nearly running the table (which I do not disagree with), the combined probabilities of a Clinton win in Iowa and an Edwards win in Iowa result in a higher overall probability of a Clinton nomination.

Shorter version: since there is a significant probability of an Edwards Iowa win (even if that probability is lower than Obama or Edwards) and am Edwards in Iowa win increases the probability of a Clinton nomination more than it does an Obama nomination, the probability of Clinton's being nominated may be greater than Obama's even if Obama has a greater chance of winning Iowa (as long as his probability of an Iowa win is not too much greater than Clinton's which it is not).

Never mind the "shorter" part.

The poll averages you list break down something like this in terms of probability of winning Iowa (need a good formula for this -- I think it may be closer than the numbers below, but for purpose of illustration . . . )

Obama 40%
Clinton 35%
Edwards 25% (case 1 below 15% plus case 2 10%)

Let's posit that each of Clinton and Obama, if they finish first in Iowa will then win via momentum. It's now Obama 40% and Clinton 35% to win it all.

But if Edwards wins Iowa the probabilities of Obama and Clinton may well be affected differently.

Case 1: IA finish: Edwards, Clinton, Obama. This is probably the more likely of the scenarios in which Edwards wins, since his winning depends more on taking votes from Obama than from Hillary given that there are two "change" candidates and one SQ candidate. Call this 15%.

Effect on ultimate nomination probability: because of institutional "firewall" in NH and $$ Clinton is hurt much less than Obama and media will label Obama as a Howard Dean whose support was just the kids, etc. Clinton and Edwards will spin the result as affirmation of experience and toughness. But Edwards would still likely lose to Clinton by the end:
Clinton: 60% chance of winning nomination still
Edwards: 20% (a quadruple of pre-Iowa nomination odds)
Obama: 20% (it's not like he will be broke and lots can happen)

Case 2: Edwards, Obama, Clinton. 10%
Subsequent nomination odds ( a real repudiation of Clinton, Obama wins NH, Edwards second there . . .)
Obama: 60%
Edwards: 20%
Clinton: 20%

Multiplying out the stuff in the Edwards wins Iowa scenario and adding to the previous totals for Obama and Hillary gives for probability of being nominated:

Clinton: 35+9+2=46%
Obama: 40+3+6=49%
Edwards: 5% (right where he is in Intrade and IEM predictive markets)

Notice four things:

1) Even though the numbers I pulled out of my ass plugged in ended up with Obama favored by 3%, Clinton gained more from an Edwards win than Obama did. The higher the probability of an Edwards win in Iowa (up to a point), the higher the probability of a Clinton coronation in the summer.

2) How sensitive these numbers are to how one translates the current poll margins into probabilities. Arguably, the probabilities of Iowa victory are closer to Obama 38, Clinton 35, Edwards 27. If so, Clinton is more likely to win the nomination (will leave the math for the reader's homework). Without having a good model for translating the polling into probability when the race is this tight it is hard to have a predictive model at all.

3) How sensitive the prediction is to the probability of finish order in the case where Edwards wins.

I love Chris's work in this area and am in no way refuting his claim. I am just trying to tease out the rest of the model. Like him, I am uninterested (for the purpose of this part of the discussion) in replies about how I just don't understand how sweet (sucky) the typist's favorite (hated) candidate is.


Visit DebateScoop for political candidate debate news and analysis.

---

by: demondeac @ Thu Dec 20, 2007 at 08:01:55 AM UTC

interesting analysis (0.00 / 0)

I give Edwards a better chance to win Iowa than you do, but that's not surprising.

I think the margin of victory in Iowa and especially NH will be important. Third place where all top-tier candidates are bunched closely together is very different from third place where your share of the vote is closer to the fourth-place candidate than it is to the winners.

Join the Iowa progressive community at Bleeding Heartland.

---

by: desmoinesdem @ Thu Dec 20, 2007 at 14:03:26 PM UTC

[ Parent ]

Good points (0.00 / 0)

From the polling averages it is a low probability that the margins in Iowa would be large between first and second, second and third. One can costruct scenarios, but that's a whole 'nother thang. A complete model would do so. I thought about adding that to my discussion but decided to keep it a little simpler.

As to what I (or you) really believe about the probability of an Edwards win in Iowa, I think you understand and hope others do that the numbers I used were for illustration. Best I can say is it's a very tight three way race. I bought Edwards shares at 17.5 two days ago and have a sell order at 30. I sold Obama at 61 and have a buy order at 40. But I "really think" it's even closer than that -- my orders are just insurance to protect my winnings.

Thank you for your series on how caucuses work, BTW.

Visit DebateScoop for political candidate debate news and analysis.

---

by: demondeac @ Thu Dec 20, 2007 at 16:43:02 PM UTC

[ Parent ]

Huckabee is interesting (0.00 / 0)

Huckabee is interesting, I'd mostly agree with him winning.  The other factions of the Republican party have money, but they don't have people.  Only Thompson I think could pull out against him perhaps around south carolina as the others fade.

The democratic side will be more about organization in my opinion and obama seems to have a pretty good organization, but Hillary is going for increased turnout as well and her voters will be more spread out geographically. 

Personally right now I'd say Hillary wins and Huckabee wins.

---

by: Astera @ Thu Dec 20, 2007 at 12:20:43 PM UTC

My prediction too (0.00 / 0)

I agree that Obama and Huckabee are the most likely to win Iowa, and that it gives both of them a good shot at winning the entire thing.  I have trouble seeing either of them wrap it up quickly since their chief opponents are well funded.

I think that Obama has a better chance than Huckabee of riding that wave all the way to the convention.  The candidates most likely to drop out of the race if Obama is doing well are quite likely to prefer him to Clinton.  Huckabee is not so lucky, every time one of his opponents drops out he will have to fight with whoever his chief rival is for the votes of their supporters.

My job is not to represent Washington to you, but to represent you to Washington- Obama
Philly for Obama

---

by: Luam @ Thu Dec 20, 2007 at 12:25:34 PM UTC

I don't know (4.00 / 1)

what Edwards supporters would do if he were out of the race before February 5. I am not convinced that in the non-blog world, more would go to Obama than to Clinton.

Join the Iowa progressive community at Bleeding Heartland.

---

by: desmoinesdem @ Thu Dec 20, 2007 at 14:04:58 PM UTC

[ Parent ]

I Thin McCain Will Win the R Nomination (0.00 / 0)

I have said this a couple times already, but I think McCain is actually going to win the R nomination.  I think the R race will play out like this: Huckabee beats Romney in Iowa; Romney gets an "Iowa dip," my newly coined phrase for what happens to front-runners who lose Iowa, and this along with Guliani pulling back in NH and McCain coming on in NH leads to a McCain win in NH.  This leads to a McCain-Huckabee show-down, which I have also been predicting for some time.  I think McCain will ultimately win this battle because the Rs are ultimately an elitist, "nominate-the-next-in-line" party and McCain will, ironically, be the candidate of the Republican elites.  Not just the economic royalists, who will throw money at McCain by the barrel, but all the right wing pundits and powers, except for certain religious conservatives, will flock to McCain.

The Democratic race is the one I think is harder to grasp because Iowa is so important and so hard to predict.  Ultimately, I agree with the blogger who said that because Clinton's position will likely marginally be improved by an Edwards win in comparison to Obama's position, plus I think she is marginally more able to recover from an Obama win then Obama will be able to recover from a Clinton win, I think Clinton is still the favorite.  I refuse to "predict" a Clinton nomination though because this possibility is, for me, too horrible for words.

---

by: Andy Katz @ Thu Dec 20, 2007 at 14:33:05 PM UTC

Iowa: 1. JE, 2. BO, 3 HC (4.00 / 1)

Just for fun, can we tease out the scenario I think is most likely in Iowa?

#1 Edwards
#2 Obama
#3 Clinton

Then what happens in NH from this current baseline?

Edwards - 16.3%
Obama - 28.3%
Clinton - 33.5%

Does Edwards get 10 from Clinton and 5 from Obama? Seems like a *very* reasonable, even conservative assumption from past history, leading here:

Edwards - 16.3% + 15 = 31.3
Obama - 28.3% - 5 = 23.3
Clinton - 33.5% - 10 = 23.5

So, granting me my initial assumption of a JE/BO/HC Iowa finish (a huge given I understand, but we're just flushing out this one scenario, remember?), doesn't NH look to be well within Edwards' reach?

---

by: GreenVTster @ Thu Dec 20, 2007 at 15:47:34 PM UTC

Am not sure Edwards actually has to win NH (0.00 / 0)

A close second might be enough. I think he may even get away with second in Iowa if Hillary is in third. The Repubs look like they have narrowed it down to Huckabee, Romney and McCain. I think Huck may yet lose Iowa to Mitt be even if he manages to hold on the GOP bigwigs will rally round whoever wins NH and go nuclear on Huckabee. I think we are looking at Hillary v Romney but I reserve the right to change my mind at least three more times before January 3!

---

by: conspiracy @ Thu Dec 20, 2007 at 16:15:38 PM UTC