Why the disparity between national and state polls?

[Doosh]

Seems like the divide is growing bigger every day. National polls show Bush tied or up, sometimes by a little, sometimes by ridiculous totals. State polls show Kerry doing great in crucial swing states. He's just about sealed up Pennsylvania and Michigan, made a comeback in Wisconsin. The most recent polls show him putting away NJ and inching ahead in both Ohio and Florida. So how can he be down to Bush nationally? I can't see this happening unless Bush beats Kerry 80-20 in red states and Kerry wins every Gore state + Florida/Ohio by 3-4 points each.

[displacedtexan]

1. 2-4% more idiots in red states responding to polls...

inflate the freak monkey's national average.

My advice: Stop paying attention to national polls.

[tedoll78]

2. Hillbilly redneck states.

Seriously though, I do think that many of the deep red states are a bit redder this time around. Much of the South doesn't like a "Massachusetts libr'l," for instance.

Meanwhile, Kerry will probably win California and New York, but not with margins as large as Gore did. These votes will take away from his numbers nationally.

And in swing states, Kerry, the Dems, the 527s.. they've all been working sooo hard to expose the Chimp for what he is - which is why Kerry is so competetive in them.

Of course, the national polls could be overwhelmingly wrong, just as they were in 2000, when 39 of the 43 pre-Election Day polls showed Bush winning popularly. In which case, this bodes even better for us.

[David__77]

4. Actually...

...I do think that Kerry will win California by 12 points this time. I would be surprised if it was under ten points. Barbara Boxer is doing wonderfully. The right is not mobilized here and Schwarzenegger's taking a pass, for the most part, on partisan activity, instead focusing on some Indian gaming initiatives.

New York, I cannot say, but Survey USA, who I generally find suspect, have the margin at only a couple points below the 25 that Gore won by.

[I Lean Left]

16. Funny you mention the polls in 2000

I remember listening to a talk show the night before the 2000 election, and the host openly wondered if the country would accept Gore winning the electoral vote but losing the popular vote. That's how off the national polls were and how close the state polls were.

I want to win both, but there is something almost poetic about giving it right back.

[Weembo]

3. skew

For a good analysis of this phenomenon, got to www.swingstateproject.com and look at one of the two latest threads

[Awsi Dooger]

5. State polls always lag changes in national preference

By several days up to a week.

We need to wait until the weekend or early next week to evaluate these state polls. Hopefully they will see no indication the Bush gains in national polls are legit.

[Euphen]

7. Usually there's a lag because the state polls are conducted . . .

. . . and released after the national polls, and it takes longer to get a clear picture of where each state stands. In this case, however, we're seeing state polls that show the race tied with Kerry gaining conducted at the same time as polls that show Bush with an average of a 3-4 point lead.

[tedoll78]

9. Bingo.

At this late point in the race, we're seeing things as they stand on the ground. No delay.

[Awsi Dooger]

10. Nope, even state polls released at the same time lag national polls

I studied that in both '96 and '00, charting national and state polls via Excel. There was a definite lag, virtually in every state.

If the national poll consensus changed by at least 2-3 points, that number would show up in state polls several days to a week later, even if state polls released on the same day as the national numbers had not indicated a shift. Perhaps there is a follow-the-leader mentality among the soft or undecided voters.

[lancdem]

13. Most national polls have the race tied, actually

[lancdem]

14. Most national polls show Kerry gaining, too

It'll be interesting to see what the Pew poll has tomorrow.

[fugop]

15. Bush gain?

What Bush gains? Aren't the gains already eroding again? Seems to me that pundits were crowing about how the debates didn't matter and Bush was back to being slightly ahead a few days back, but today at least a few of the nationlas seemed to be trending back to Kerry. NBC, Zogby, Rasmussen all seem to have showed Bush's gains slipping.

[Awsi Dooger]

19. I saw the head of the Washington Post poll interviewed today

He said their sample has seen a one point Bush gain per day average since the end of last week. I have no idea if that's legit or representative.

My Excel tracking of the national polls does show slight gains for Bush since the third debate. And I don't even include Gallup. Right now I'm waiting for ARG and PEW, two firms I respect. They have not updated lately.

Zogby is potentially volatile because he comes up with bigger daily swings than others, and begins assigning leaners and undecideds late in the race. According to a post here the other night, Zogby tracked a 49.6 to 39.1 Kerry advantage one day last weekend. Once that figure is dumped from the 3-day average, Bush could regain the lead. I hope that's wrong but I want to be prepared for it so I don't overreact. Still two full weeks.

[Nicholas_J]

6. Becasue when they do national polls

The do not weight them by polling more poeople from larger states than from smaller states. They simply do an even slit in the number of people they call in each state. SO in a poll with 1000 people, they are calling TWENTY pople in each state.

This means that the smaller less populated and largely red states get as many telephone calls as an enormous blue state like New York or California. Sinced Kerry's leads come from the states wit the highest populations, the pecentage of people who support Kerry are underrepresented in national polls, while in the state polls, the number of people polled is a much larger sample. When you looks at Red state, only one extremely large state falls into this catagory. Texas. Kerry is ahead in the two largest states with more electoral votes, New York and California. THe next largest state, Florida, is running neck and neck with a small lerad for Kerry, in the lead by one point, withing the margin of error for that state, though no matter what, even though Kerry leads a little in Florida, most of the time this state is thrown towards Bush because he won it in 2000.

While this would not make a differnce in the polls for the party nomination (democrats runniing against democrat for the nomination in ared state wont make all that much of a difference) ithas the effect of skewing the polls to whoever has the lead in the most states, regardless of the actual population and electoral vote values in those states.

[HeldsBelds]

8. no

that's not correct.

National polls are made up of random digit dialing. They don't equally weight the states.

[DemocratSinceBirth]

11. With Random Digit Dialing You Would Still Get More Respondents

from big states because they form a larger part of the pool...

[Nicholas_J]

12. yup

most pollster including Gallop use the most unsophisticated methods of doing polls, like choosing numbers from the 2000 Phone disk sets and the randomly choose a set number of people from each state. This is why you get someone like Gallop regularly polling 39 percent republicans, 30 percent democrats and the rest independent. Gallop uses a list of voters by voter registration from each state. Then they select out fixrd number of people they are going to call, take 39 percent of that number, and divide by 50, and then select from the list of registrered republicans. Then they do the same fot 30 percent that will be democrats, and then the same for indies

A simple random dialing would not continually end up with the same fixed percentage of democrat, republican and indie that they predetermine at the beginning of the year.
THey get listings of voters from the division of electionsfrom each state, Zogby and ARG on the other hand use a national phone list, but even they do not randonly select from that one list, but do a division by state. Only Zogby and ARG have not set up a model of anticipated percentage of voters per party.

SUSA, Gallop, and the pollsters for Washinton Post/ABC. CNN, NBC, and CBS all use a preselected percentage of voters by party, using a model determined at the beginning of the campaign season. The have all for some reason determined that this year more Republicans will vote than democrats, so they predetermine a number of people to call to get their pre set percentages.

This is why Gallop and SUSA are both oversampling Republicans. The other two are as well but not to the same degree. If you lookat the specifics of the SUSA polls, you will see that they are polling a significant percent more Republicans than democrats, and that this is done by model.
Simple random calling does not fit the models and cannot reliably get the predetermined percentages they have established. The percentages for every Gallop poll bhave been set at 39 percent Republican , 30 percent Democrat. Every poll they have done uses that preset percentage breakdown. So dies SUSA.



.

If they used a random. such as you are suggesting, they would not get the pre-sepected percentages of Republicans and Democrats that they have decided are the proportion in which they have calculated will vote in the election this year based on the percentages that voted in the prior presidential election years. They pre-select the percentage of republican

[nicktom]

17. Thanks for the insight, I hope you are right.

By the way my oldest son is named Nicholas, 20 years old and primed for an up and coming draft if * is re-selected.

[Nicholas_J]

20. I have read articles that indicate

that some of the pollsters like Gallop and the pollster who does some of the polls for ABC (some PA pollster who's website has a section on political polling that still reads "UNDER CONSTRCTION" are having to make between 30 and 70 calls until they can find people who either want to answer them, know something about what they are talking about and then fit the model that they have decided is going to be the proportion of voters they claim will be voting this year (you know oversampling Republicans).

Its gotten to the point that pollsters are doing polls to see if people beleive the results of the presidential poll. Of course, Gallop says most do, and independent pollsters like ARG come up with most dont.

[Nicholas_J]

21. My nephew was almost talked into the Navy

by a recruiter who got his hooks into the kid just as he got out of high school. My sister and her ex-husband couldnt talk him out of it so they got me to talk to him and I got a couple of people who just got back from Iraq (including my doctor who was an officer and just got back) and the idea went RIGHT out of his head. I understand the recruiter wanted to talk to ME once the kid told him why he decided not to join up.

A couple of pix from from press other than the American press helped as well

[Mayberry Machiavelli]

18. But then shouldn't the pollsters ask numbers of people proportional

to the population size of the state to get a true "national" sample?

I.E. if NY is twice as big as Michigan, sample 20 voters in NY to 10 in Michigan. If NY is 200 times as big as Montana, ask 200 NYers to 1 Montanan, etc...

What you are describing is an automatic GOP bias to "national" polling vis a vis the election, with the way the party alignments currently are.