Deinonychus antirrhopus

I think you make of it the same thing you do of opinion polls about lawyers. Everybody hates lawyers, but they're okay with their lawyer. A generic House poll reveals little, since it's actually 435 races with no voter overlap between them, and often turn as much on individual personalities as party affiliation. I'd love to see a GOP meltdown, but given the dynamics of incumbency, tide changes in that body are pretty hard to come by.

This poll is absolutely all over the map. Look at the 3-way race. Bush has a larger advantage with Republicans than Kerry has with Democrats, AND he is winning with independents. Yet Kerry is up by 7 and Congressional Democrats are up by 19?? It seems quite obvious to me that there is a massive oversampling of Democrats in this poll. If I were a polling director, and I saw either Congressional party up by 19, I'd trash the poll and do it again. This poll is a joke.

Tangent Topic Poll: What do you consider to be the most credible polling organization?

I think I saw somewhere that the LA Times poll had 13% more Democrats than Republicans. And yet, Kerry only leads by 7%?

Not correctly for likely voters is a very old, long exposed, pro-Democrat dodge. That the LA Times is resorting to it shows how low their credibility has fallen.

Not 'correcting' for 'likely voters' does favor Democrats numbers, as their part of the 'registered voter' group is less likely to vote, given the measures by which that is quantified.

However, it is not a pro-Democrat dodge, but rather, the way the major polls have ALWAYS done the voter screen-- they use the 'registered voter' screen for polling before about Labor Day, and then switch to the 'likely voter' screen once the election season hits high gear.

It is Zogby's methodology, in his constant use of the 'likely voter' screen, that sets him apart from other major polling organizations. Not that they don't turn to that method eventually themselves, but that they haven't, historically, until far later in the election season.

Zogby's take on his method is that it favors the GOP side by a half dozen points or a little less.

If we are not professional pollsters, how to choose between these methods? Well, by universal practice, all agree that once the electorate is fully engaged, paying attention, down the home stretch, the 'likely voter' screen is preferable, and more accurate. Why wouldn't it be equally preferable and more accurate early?

The theory is that early in the process, people aren't paying much attention to the issues that would motivate them to vote. Some of the 'likely voter' screens are a) did you vote last time, and b) do you plan to vote this time (and they need to say they're registered).

Obviously, SOME who are now not paying attention enough to think they should vote WILL decide to do so, as they get more information. It seems the traditional pollsters' methodology finds the early inchoate electorate as well modeled in the 'registered voter' screen as in the 'likely voter' screen (at least, given the additional money and effort requred to get a significant sample size of likely voters).

The only poll of significance is on November 4th.

Fred, that is not completely true, or both sides wouldn't spend millions of dollars on constant polling, as they do.

But, of course, these polls do not pretend to predict what will happen 5 months out from here, but only what would happen 'if the election were held today.' That becomes relevant as 'today' gets closer to early November, and for the trending that polling shows.

Personally, I like the approach taken by the folks at realclearpolitics.com Report on a whole BUNCH of polls. Include (separately) both nation-wide and regional/local polls.

If 8 polls indicate the President's support is slipping, and they're all in roughly the same range, then whether you think it's slipping by 2% or 4$ matters less than whether the President's support is slipping.

Similarly, as noted above, local politics means that a poll on generic House votes matters a LOT less than state-polls (often by state papers) on state races. So, if you wanna take a guess at who'll own the House come November, it's probably more useful to look at a bunch of local polling reports (and w/ blogs, that's a lot easier than trying to parse the WaPo or LAT's reporting).

I suspect that the respective campaigns' polls do more than just "if the election were held tomorrow" type stuff. I think they're trying to get a much better guess as to where they're weak (and therefore have to spend money). And I'll bet that they're much more defined, thanks to GIS systems.

Just a thought....

I have long since stopped considering the LA Times to be a credible news source, so I consider this poll to be as interesting as an alien abduction story in a checkout-line tabloid. Proper Bayesian updating of beliefs assigns a very large prior probability to the likelihood that the LA Times is distorting the facts.

This is the same paper which wanted to keep Grayout?

And why should we believe what they say?

The poll results are not unimportant, even though they are highly unreliable. But what is more important is that the Time's publication of the polls is likely to affect the election results, by telling Republicans that their votes are useless while telling Democrats that their votes will make them part of the winning team. It's a subtle way of editorializing on what the result of the election should be while maintaining the fiction that this is a scientific, objective conclusion.

It reminds me of a former mentor who had a very cynical attitude towards politics (and just about everying else, except himself.) Once, when confronted by the fact that the legality one of his pet projects was highly questionable, his attitude was "I will _make_ it legal." Though I learned a great deal from him, his cynicism eventually became so irritating that I felt like pushing him off a ledge.

In a sea of polls, so many out there, and so many to come, it is ridiculous to argue any paper's purpose in putting out the results of a poll taken in June (or May?) is done to change the election results in November. A) it is likely impossible for any paper's polling to have that effect, and B) any such effect would only happen with polls closer to election date, and C) presumably, if the LA Times would influence any, it would be California voters, and California is NOT a battleground state by anyone's calculus, but already a safe Democratic electoral college pickup.

Not only is the House elected in 435 different elections, but Presidents are elected by 50 different elections.

Polls have a lot of built-in problems. Dems are supposed to be more likely to answer the phone, but by what margin? Polls typically ask 30-60 questions with no control over the interaction between questions. National results are not accurate since electors are determined by state. Any genuine statistician would be embarrassed by polls.

They're better than no information, but not by much. There was a poll a while back that showed Bush within 1 point of Kerry in California.

I like the Iowa Electronic Markets. People may do bad polling for editorial purposes, but when you have real money on the line you tend to pay attention.

I like the Iowa Electronic Markets. People may do bad polling for editorial purposes, but when you have real money on the line you tend to pay attention.

Plus I think the information set is larger when you look at the people involved in such markets as a whole. It isn't that they simple average over all media polls, it is that they look at all such polls and additional information as well.

Good point on the CA poll Ron. I recall hearing about that one as well. It was probably wrong too. CA may have moved a couple of ticks to the right with the screw ups of Davis and his ouster, but, it is still only a couple of ticks (if that...Davis was insanely unpopular).

The pollster's take a bad experimental design (and they know it's bad, I've emailed questions to them), they do questionable randomization, and I doubt they test their questions for bias, and then they have the gall to quote an error percentage as if they got everything right.

Irritates me no end.

I'm at least positive that the CA poll was wrong, maybe more than positive (as soon as I can figure out how to do that).

I'd throw out a joke, but I read some PJ O'Roarke over the weekend (I hadn't ever read him before, what a miss). I'd feel ashamed to attempt humor after that. I bet the Dems really hate that guy.