The 2:1 rule debunked

[Silent Hunter]

http://www.mysterypollster.com/main/2004/10/more_on_the_inc.html

Comments please.

[12th Doctor]

I posted something about this a long time ago.  The undecideds in a Presidential election acctually rarely break for the challenger.  The data that suggests so is acctually from Congressional races.

In 1976, 1980, 1984, 1992 and 2000, the late deciders broke for the incumbent party.

[Democratic Hawk]

The 2:1 vote for the challenger rule is far from absolute. Sometimes it will favour the challenger, other times the incumbent.

I'm not sure how I'll spit the undecideds once I make my state-by-state predictions. I'm consdering 60:40 Kerry in blue states and likewise for Bush in the red states.

Dave

[millwx]

http://www.mysterypollster.com/main/2004/10/more_on_the_inc.html

Comments please.

Well, that article certainly doesn't debunk it (in fact, the "86%" discussion implies that it's more like 5:1!), but it does clarify things.  On the one hand, the "rule" does not exist when considering polling data well before the election.  And those are the polls, with higher undecieds, in which there would be more impact if this held.  But it simply doesn't.  On the other hand, there've been a lot of folks on this board who simply say the "rule" doesn't exist at all.  This isn't correct either... this article presents some evidence, as does the article on PollingReport, and even our very own Vorlon showed that, while the data is inconsistent, there is some evidence that it exists.  And some of the data in the PollingReport article, this article, and all of it in the Vorlon's analysis did include Presidential elections.  So, this isn't simply due to Congressional elections.  One bit of "corrputing" data amongst those with that opinion is that they insist on looking at the incumbent party.  For example, they'll look at 2000 and say that it broke for the incumbent (Gore).  That is simply not how this "rule" has been analyzed or how it works.

Basically, I think what it amounts to is that Kerry will pick up the undecideds 2:1, but we're talking about only the true undecideds just before the election.  So, we're probably looking at no more than about 3-5% of the electorate.  That means that Kerry will probably gain only 1%, maybe 2%, on Bush simply based on this.  (And I should add that even if one doesn't believe this "rule", internals of undecideds right now, does indicate that they may well break to Kerry... so, real or not, it should hold this year.)  Frankly, IMO, this won't be enough.  Bush looks like he's up by about 3-5% right now; so, he'll likely win by 2-4%... enough to ensure an Electoral Vote win.  So, it may well be a moot point anyway.

[elcorazon]

http://www.mysterypollster.com/main/2004/10/more_on_the_inc.html

Comments please.

Well, that article certainly doesn't debunk it (in fact, the "86%" discussion implies that it's more like 5:1!), but it does clarify things.  On the one hand, the "rule" does not exist when considering polling data well before the election.  And those are the polls, with higher undecieds, in which there would be more impact if this held.  But it simply doesn't.  On the other hand, there've been a lot of folks on this board who simply say the "rule" doesn't exist at all.  This isn't correct either... this article presents some evidence, as does the article on PollingReport, and even our very own Vorlon showed that, while the data is inconsistent, there is some evidence that it exists.  And some of the data in the PollingReport article, this article, and all of it in the Vorlon's analysis did include Presidential elections.  So, this isn't simply due to Congressional elections.  One bit of "corrputing" data amongst those with that opinion is that they insist on looking at the incumbent party.  For example, they'll look at 2000 and say that it broke for the incumbent (Gore).  That is simply not how this "rule" has been analyzed or how it works.

Basically, I think what it amounts to is that Kerry will pick up the undecideds 2:1, but we're talking about only the true undecideds just before the election.  So, we're probably looking at no more than about 3-5% of the electorate.  That means that Kerry will probably gain only 1%, maybe 2%, on Bush simply based on this.  (And I should add that even if one doesn't believe this "rule", internals of undecideds right now, does indicate that they may well break to Kerry... so, real or not, it should hold this year.)  Frankly, IMO, this won't be enough.  Bush looks like he's up by about 3-5% right now; so, he'll likely win by 2-4%... enough to ensure an Electoral Vote win.  So, it may well be a moot point anyway.

nice post.  not sure I agree with the conclusion; it depends on how close the election is right now.  If Bush is up 48%-46%, that could mean a dead heat; if bush is up 49%-46%, then Bush will still be ahead after the "break".  I'm hoping it's close enough to make Nov. 2 interesting.  One thing is for sure, barring some unforseen event, Kerry will NOT break 50%.

[AuH2O]

My whole rationale for being confident in Bush's victory is that Kerry won't dominate the 'undecideds.'

Keep in mind, these are actual people, not numbers. What happened in 1976 is no basis at all to make some kind of rule about voter behavior.

The reality is that late deciders see the same candidates as earlier people, they just have not paid sufficient attention or decided what their views are. So, Kerry is still Kerry regardless of when you decide (and vice-versa).

Kerry is not popular or well-liked- his favorable ratings are absolutely terrible... historically bad. From day 1, it was my view that his lack of popularity would be most damaging with regard to late-deciding 'swing' type voters, who obviously have not clearly prioritized the issues in the race (otherwise they would know which way they were leaning).

In other words, I just don't think Kerry's vote is going to show-- he'll do better than he should, because Bush is weak, but I capped him at 47% 9 months ago or so and I believe that prediction remains accurate.