Mason-Dixon: Allen (R) increases lead over Webb (D)

[Alcon]

I knew that once mason-dixon came out, Webb will be even down further. But I think this poll is understated I think that Rasmussen is right, he is only down by 11. And also, everyone says that Mason dixon is the most accurate, they didn't fair that good in the 1990's where they predicted in 1994 Senate race right in TN.

We are going by polling results for a 1994 Senate race to judge a polling company's performance a decade later?

I'd rather go by 2004.

[Alcon]

All I am saying is that you have to average all the polls together, you don't just take one poll company and then think that they will rule the whole universe.  Also, they were wrong in MN in 2004.

Mason-Dixon wasn't perfect in 2004, but I heavily biased my projections toward it and ended up with the highest score out of nearly 2,000 predictions.  Granted, I did miss Minnesota (and ME-2 - don't ask; I was dumb), but Mason-Dixon helped me nail virtually everything else.  And I'm pretty stupid.

I'm not saying that you take one poll company and rule the entire universe, but when you have:

1. Rasmussen releasing odd one-day polls with equally odd results.
2. Zogby doing whatever it is that he does to generate polling numbers (I think it's legal as long as they do it in Guatemala).
3. Traditionally decent pollsters like Survey USA, Strategic Vision, etc., staying fairly quiet.

...a Mason-Dixon is a welcome sight.  Statistical error means that they are going to screw up every once in a while.  Nobody's perfect.  But, by and large, they have never done anything dumb since 2000, and they are really the only pollster I can say that about.

Now, I will also point out that averaging every poll with no consideration of relative release duration and pollster quality is no more reliable than just taking Mason-Dixon's word for it.

[Alcon]

Don't forget ARG  and Gallup haven't released poll numbers on OH and MO Senate races and they tend to favor Dems.

ARG is another one of those solid pollsters that has been kind of quiet this year.  Same to Research 2000 (although they are always quiet - so much so that they appear more solid than they might be).

Gallup had some disturbingly random swings in 2004.  They were freakishly good and predicting the way the contest would drift in a week after their poll release, or so.  When the race was going from Kerry +3 to a Push, they'd have Bush +4-5.  It didn't make any sense, but at least it was interesting.  Still, I'm wary.  I like Gallup, and always have, but results like those just indicate that there are issues.

[Alcon]

Alcon, all I am saying, is that you should consider the conservative ones as well as the liberal ones. And don't forget, eventhough Gallup got the states wrong, they predicted Bush would win 274 electorial votes.

Oftentimes, it seems that the polls that lean one way shift.  It's all about choosing a sample size, and keeping it consistent.  Some polls assumed even turnout in 2004; others assumed a Dem advantage.  The former turned out better than the latter.  I truly forget which was which.

I haven't seen any real bias in Mason-Dixon.  Beyond that, I usually shift Strategic Vision a point to the left - there seems to be a consistent error there.  Sadly, especially with Senate polls, it's hard to tell until it's too late to tell.

[Alcon]

No, I didn't say that I said that you should be weary of polls that constantly show GOP ahead, just like conserv are weary of Zogby polls that show Dems ahead.

I wouldn't necessarily say Zogby has a heavy Dem bias.  He's just so often wrong wildly, and about two-thirds of the time when he is, it's toward the Democrats.  Overall, it gives him the largest bias, but hat doesn't mean he doesn't release excessively GOP-friendly polls.

What evidence do you have that Mason-Dixon underestimates Democratic support?  As Sam said, summer polling tends to overestimate Dems.  Wait until September, and Mason-Dixon will become oddly Dem-leaning, before leveling off before the election.

[Alcon]

I said the sample is small that understates Democratic support. Most liberal pollsters uses a 1000 polls. That's what I am saying.

Liberal pollsters?  Eh?

625 is perfectly fine.  There is a point of diminishing returns, and even factoring in MoE, the absolute closest this race could be is Allen 44%, Webb 36%, unless Mason-Dixon missed MoE, which has never happened.

[Alcon]

Well I am going to stick with my liberal polls that show Dems leading and have large samples instead of using small samples.

625 is not a small sample, despite what you are saying.  And sample size adjustment is what MoE considerations and for.  Even considering the MoE of this poll and being overtly friendly to the Democrats, the closest this is, is Allen +8.

The range is Allen +8-24.

[Alcon]

Well, Alcon, the fact of the matter is that they use a different sample of likely voters than Zogby does or Gallup and that is a fact. The only state that is for sure is PA and even that I can't predict.

The only state that is for sure in what way is Pennsylvania?  And, obviously, every poll company has different samples.  Otherwise they would all get virtually identical results.  What is your point?

[Alcon]

The mason-dixon poll company says it is registered voters most likely to vote. So, they do both.

That is not correct. "Registered voters most likely to vote" means "likely voters."

And I still do not see how you claim that Mason-Dixon has a "conservative lean."  Just because it is giving more conservative results during the summer than the polling average does not mean it is conservative.  It could mean:

1. They are better for adjusting for the issues that affect summer polling results.
2. The other pollsters are too far to the left in their samples.

Few dispute that the election turnout among Democrats and Republicans will probably be fairly even nationwide, as it was in 2004; that's not the point of contention, necessarily.

[Alcon]

You keep saying that it is liberal, they have never had Brown ahead so I don't think that they will. And they never had Kerry lead in summer or fall in OH.  Look, Mason-Dixon oversamples republicans more than Dems, they simply don't have the liberal sample that Zogby or Gallup. When it selects likely voters, they select a more conservative sample than liberal ones.  And 625 is too small of a sample. The sample of likely voters is the key. And lastly, they predicted more Republicans to turn out than most other pollsters had. Most pollsters had it 35R and they had it 37R.

I keep saying what is liberal?  Where did I say this?  What are you talking about?

Why is 625 too small of a sample?  Midterms tend to have higher Republican turnout, marginally, in most cases, I believe.  Again, just because they have a higher percentage of Republicans than average does not mean they are wrong, or right-leaning, or whatever.

I think you are investing too much interest in pollster's individual leanings.  As we get into October, you will see that all pollsters will start floating together. You will also see that Mason-Dixon will start leaning libreral in September.  That is, unless you contest that summer polling doesn't lean Dem and early fall polling lean GOP, which is just untrue.

[Alcon]

I said that you keep thinking that Masondixon is liberal, it isn't they never had Kerry ahead of Bush in Oh and Brown ahead of DeWine. And the pool size of likely voters is different than most other pollsters. They are an independent poll company but they have a conserv slant and you aren't going to change my mine on that. And also, Survey USA and Columbus Dispatch does mail in ballots and they do phone samples. There is a difference. They poll more republicans and liberal poll more Dems and we wll see which one is correct at the ballot box. This time, maybe Mason-Dixon will be on the losing side like it was in 1992 and 1994 and just like they predicted FL and AZ going to the Republicans in 1996.

Bush won Ohio.  There is nothing wrong with Mason-Dixon's numbers in Ohio.  They nailed the final result, and weren't insanely far to the right at any point.  Just because other pollsters had Kerry ahead does not mean he ever actually was during times when M-D polled.  That is not proof that Mason-Dixon is more right-leaning than the results; that is proof that Mason-Dixon is more right-leaning than most pollsters, up until the end of the election!  We have no way of knowing who was "right" in September or October.

[Alcon]

Bush won OH because M-D correctly predicted the outcome of more republican voters were going to turn out than Gallup or Zogby.

Look at the Gallup poll prior.  That wasn't their issue.  And Zogby has completely unrelated issues.

Their ability to predict that sort of thing seems like a good thing to me.

[Alcon]

Of course you don't see it my way because you are not a Democrat. If m-d wasn't so conserv how come conserv listen to them and liberals don't, that should tell you right there.

Sorry to disappoint you, but I was a liberal Democrat in 2004, when I got the top score thanks to Mason-Dixon.

[Alcon]

Look, polls are snapshots in time, and no matter what a pollster might say, turnout is the key. And no one can predict for sure how the turnout will be and turnout can be the key to Democratic success this fall because they are upset with Bush. As Howard Dean says "turnout can overcome any deficit." He was saying that in reference to Kerry trailing in OH on election it didn't happen but it can be true, just like everyone expected Tom Daschle to beat Thune and turnout was the key in that race on election day.

True, and Mason-Dixon has a sterling record with predicting turnout.  It almost sounds like you're saying that there is no way to tell who is right, so we might as well go with arbitrary measures (most Democratic, most people polled, smallest MoE).

[Alcon]

And Zogby was just as good like in 1996 and 2000 when he polled less republicans.

Well, in 2000, I would have trusted Zogby.  In 2004, I would have noticed that his results were getting too erratic and not trusted him.  I do not see this in Mason-Dixon.

Zogby polling fewer Republicans in 1996 and 2000 is interesting, but irrelevant as to why I should not be trusting Mason-Dixon as much as I do now.

[Alcon]

All these independents and republicans jump on the mason dixon bandwagon when they are right, but where were they when they weren't right in the 1990's.

I was a little busy being more interested in learning cursive and long division than worrying about polling mechanics.

I would not have defended Mason-Dixon inthe 1990s.