2004 User Predictions - Discussion

[millwx]

Just randomly chiming in here after seeing lots of folks' predictions.  Just my two cents, but I think a lot of people are "missing" some standard, well known polling issues...

1) Third party candidates poll roughly double what they'll actually get.  The other half migrate to the major candidate most closely matching their's... In this case, about 50% of the polled Nader support will go to Kerry.

2) Undecideds go heavily (about 70-80%) to the challenger.

This means that almost every state polling even or with Bush up by 1-2% is actually, likely, leaning Kerry.  Zogby said some of this (more simply) on his site yesterday, and he's absolutely right.  This is Kerry's race to lose.  Due to these polling "issues" Kerry is likely ahead in all of the Gore states plus NH, OH, FL and AR... maybe more!  Obviously, many states are close; Bush coulde easily win... and by a lot... with only a slight surge.  But right now, Kerry is ahead... significantly... if one interprets the poll numbers properly.

[millwx]

Having any idea who is going to win this thing in May is like sitting back and trying to predict who is going to the World Series. You can make a lot of good guesses but will probably be wrong in the end. We still have conventions and debates yet to go! Thats a lot of stuff still a long time off.

While I completely agree with that notion, the fact is, many people ARE trying to make such a prediction.  So, I was just trying to point out that if anyone has any hope at all of doing that, they need to at least understand how to interpret the polls. (That comes off as rather "holier than thou"; sorry, it is not intended to... I'm merely trying to point out, from a lot of poll-parsing I've done, as well as taking advise from the polling experts themselves, that there is more to polling data than the mere straight-up numbers.)

[millwx]

I like to look at state polls more then one big national poll. Yes it looks bad for Bush right now but I know that will change as the summer wears on. I really don't put faith in polls as I don't believe the sample enough people. In this modern age we could easily sample 500,000 registered voters. Just my opinion, it may not be feasible I don't know.

BTW, what part of MD are you from? I grew up in N.E. Baltimore City. Franklin Square Hospital is where I was born.

I agree (about state polls).  I think national polls are important in that if someone has a wide lead, then, don't bother parsing the state data... that candidate will win.  But when it's close (like this), the state polls are the key.

As for Bush improving over the summer, I'm not so sure.  News continues to be bad for him (except for some jobs data), and I see no reason why it won't continue.  Meanwhile, Kerry hasn't even really begun his campaign (o.k., well, he JUST has - as he's launched an ad buy).  By the way, I'm not spouting this as just some personal preference.  Frankly, I don't like either candidate.

Anyway, I agree that it's early and much can change.  There are so many tight battleground states that any significant improvement by either candidate could easily result in a landslide.  To my original point, though, to me it doesn't matter whether the election is 6 months or 6 days away... if one uses polling data to make their prediction, they should use it with the best knowledge of the polling biases/errors/tendencies possible.

Incidentally, I'm originally from Massachusetts (I'm a "Northeast Republican"... which makes me more libertarian or independent in the rest of the country :-) ), but now living in SoMD (PG Co.).  I did recently live in Baltimore City, though (north of Canton/Fells Point).

[millwx]

I tend to agree with you, although I'll reserve that opinion until I hear Vorlon's view of it.

Just wanted to pop in again here, because apparently (it was mentioned on the MI poll thread) Vorlon (I presume he's your resident poll expert?  I've not been around here long enough), has confirmed my statement (not personally said I was right... but he has stated something similar... I didn't see such a post by Vorlon, but someone was referencing him)...

The challenger gets a vast majority of the votes.

Though I admit it's a very questionable thing to do due to methodologies, I can also tell you that if you take the six most recent national polls and average the three with the largest pool of undecideds versus the three with the smallest pool of undecideds you'll see that, indeed, the undecideds break better than 2:1 for Kerry.  I'll admit it's a bit shady to be averaging polls like that, but it does indicate that the 2:1 rule does hold.

I mention this because a lot of Bush supporters on this thread and the MI poll thread are disregarding this "rule", for a number of reasons.  I'm not being pro-Kerry here (I don't have much of an axe to grind, I don't like either one), I'm just being realistic.  The Dems appear correct in flaunting this axiom around.  The "resident expert" has confirmed this, and current poll averages IMPLY the same (though, again, I'll say that poll averaging like that is a bit shady).

So, with about 5-8% undecided in most states (more in some), Bush needs to be about 2-4% in the lead for it to really even be a dead heat.  If I could figure out how to post my prediction here (I'm a bit of a techno-idiot), I would.  It's straight from the poll numbers, but using the 2:1 interpretation and tipping any ties to Kerry, due to the trends plus the Nader factor (some Naderites will likely go for Kerry at the last minute).  You'd see that it's far more Kerry-colored than a map using the poll numbers straight-up, with no interpretation.

Anyway, sorry for being so long-winded... I don't post often, so needed to get all my thoughts out at once!  :-)  Just wanted to chime in again that there is some validation on the 2:1 (or more!) issue I raised earlier.