Interesting Poll @ Princeton

[sjdnb]

I stumbled on this poll and wanted to share it. It appeared to me to use sounder methodology than many others I've looked at.

http://synapse.princeton.edu/~sam/pollcalc.html

[snowbear]

1. Whoa baby!

A sound methodology and a very nice forecast ;)

Go Princeton.. :)

[prodigal_green]

2. thanks, this is cool

I like his explanations. Very thorough but easy for the layperson to read.

[DeepModem Mom]

3. Seems I recall that Prof. Wang's analyses are respected --

wonder why we don't hear more about them?

[humanbeing]

4. I've posted that on DU

a couple of times, but the scroll knocks posts back so fast, nearly nothing sticks.

[Awsi Dooger]

5. True, statistically-oriented posts tend to have short front page life

If you want longevity and dozens of replies, something like "Judy Woodruff sucks!" works every time.

Sam Wong is terrific. I've stolen many ideas from him over the last couple of years and updated my Excel election spreadsheets.

My theories and numbers give Kerry an advantage, but not to the degree of Sam Wong or TruthIsAll. I have charted state polls for 8 years and come up with a PAN, or Partisan Adjustment Number, for state polls. If state polls in New York, for example, consistently underestimate Democratic strength, I'm not going to ignore that. I have a PAN adjustment for each state and apply it to my average of the reliable state polls. Let's say New Hampshire has a PAN of +1.2 red. If the polls suggest Kerry leads by 2.0 points, my PAN amends that to +.8. This is before assigning undecideds. Some of the polls Wong quotes, like Survey USA and Stretegic Vision, are among the worst and I don't use them if others are available.

The PAN is similar to a pointspread in football. Many of the contested states this year have red PAN, meaning I believe those polls slightly overstate Kerry's status.

The trickiest aspect is assigning undecideds. There are simply not enough incumbent presidential races to make a completely reliable forecast. Anyone who tells you otherwise is a fool. I like to look at all incumbent races in terms of similarity to others in the expected range of outcome. Similar to sports betting, where I would never be ignorant enough to evaluate a game with a 14 point spread the same as a pick'em contest. In looking at dozens of state/local incumbent races in the 50/50 range, it's obvious the undecideds do not break as sharply to the challenger as in a lopsided race, either in favor of the challenger or incumbent. My Excel estimation is Kerry will receive 62-63% of the undecideds nationwide.

I question the standard practice of Wong and other forecasters who distribute undecideds by the same percentage in every state. That ignores partisanship qualities of the state. Undecideds do not break sharply against the grain of the state. We would have to be pretty silly, for example, to look at a Utah poll with a Democratic challenger tied at 45-45 and expect him to get 60-75% of the undecideds.

[humanbeing]

6. do you have your own site for this data? n/t

[DemocratSinceBirth]

9. Certain Polls Do "Suck"

but suck is a subjective term so wouldn't it be more scientific just to include all the commonly accepted polls...

[Zing Zing Zingbah]

7. That's one funky electoral college map!

It looks like it was done by Picasso. I like it.

[DemocratSinceBirth]

8. It's Funny...

You learn in school that Alaska is the largest state in the union but scaled for it's population it looks like a speck...

[Gloria]

10. kick!

eom

[crispini]

11. Thanks, good stuff.