Can someone explain how "internal" polls would be better than Gallup, RAS, Zogby, Etc....

[KansasVoter]

I can't imagine Gallup not knowing something that the internal pollesters figured out.

Maybe I am just stupid but it seems like any major polling company could do the same as the internal polling companies do.

[Thrill]

1. Can you explain why you announce your internal polls?

Because they're full of shit

[DrToast]

2. Well, look at the three you mentioned

Gallup, Rasmussen, and Zogby. They're all showing different numbers right now, so we know that at least one of those has to be more accurate than the other two, right?

Same idea with internal polling. The candidates believe their polling is better.

[ErinBerin84]

3. they must be using Rove's magic "THE math"

[mod mom]

5. aka "the diebold effect"

[abbiehoff]

4. Internal polls are better because you can poll anyone you like.

Or at least anyone who likes you.
That way, it looks as though you have a chance (when you probably don't).

[AllentownJake]

6. Candidates have different internals

based on turnout models. McCain probably released his internal showing best case scenario for himself.

[backscatter712]

10. There are polls for different purposes.

There are the public polls like Gallup's daily tracker that they throw at us plebes.

Then there are the serious polls that they use to do research - we never see those, and they use them to try to figure out which lines of attack work best.

Then there are the polls used to create a dog-and-pony show - probably a push-poll, designed to make Obama look like he eats babies, and make McCain look like the Messiah, and with numbers deliberately cooked to make him look good, and he'll show us those to try to make it look like he's doing well.

[TheCoxwain]

24. Did you attend USNA? Plebe is a very Naval Academy term.

nt.

[InternalDialogue]

7. It's gotta be no more than motivation for R voters

Seems to me, if your internals showed a stunning upset, you'd just hold your cards close and enjoy the surprise on November 4.

But... if you're down in the polls and need a Surge™ of voters to change the landscape, you'd announce a bunch of bogus numbers that give hope to the hopeless and possibly turn out more R's on Tuesday than the major pollsters expect right now.

[grantcart]

8. Internal polling is a much better source of polling information for 2 reasons

internal polls are based and millions of data entries that are based on the canvassing of thousands of volunteers.


The canvassers are not randomly picking a few thousand voters but are going door to door in key bellweather precincts and counties.

they are also rating responses on a 5 point scale from Strongly support to strongly against.


These databases are updated daily as more phone and door to door canvassing is done and analysts can then watch an entire state and key areas and compare with past performance based on previous elections.


Pollsters have to take a representative sampling for cost effectiveness -- internal numbers are based on an encyclopedic database that is constantly being updated by an army of canvassers.

The campaign can then retain polling firms to poll special areas to find out more about the issues that are moving people one way or another.


Its not unusual for close campaigns to have almost the identical internal polling and sometimes stumping the 'professional pollsters'.

[maseman]

13. This is partially true

Using door to door canvassers to do polling is with mixed results. Yes you can get a better in-depth look at what is going on. But the people canvassing are not going to random households. They are going to households where they are either trying to GOTV with Rs or former determined moderates. They most likely are going to no new voters or Dems households.

Using canvassing people's data is more of a focus group than a poll. It is helpful for determining if the message is good or not, if ads are working or not, if certain buzz words are working or not.

In addition using volunteers to do polling is tough because they simply cannot train these people to do "polling." People who work at Zogby, Gallop, etc. have in-depth interviewing experience.

[AllentownJake]

16. McCain doesn't have the volunteer force

in some states he has had to resort to pay for canvassers. McCain's data probably isn't that good.

[RichardRay]

17. It's always been my impression

that a well-chosen and truly random sampling of a small percentage of a population is more representative than a less random and much bigger sample. the bigger and less carefully chosen sample can actually amplify sampling error.

As to why campaigns use them: The really internal internal polls are used to look at questions the public polls don't ask. Say Barack's campaign believed that there was a high correlation between households where more than a single language is spoken and their liklihood to support Barack. Before they spend beaucoup bucks on an ad campaign aimed at those households they want to know if that's so, so they commission an internal poll. The campaign may also have developed an internal model that's better than the big public pollsters. Gallup et al are very good, but there are some pretty high powered statistics and behavior guys working for he campaigns, too.

The other kind of 'internal' poll is one done specifically for marketing reasons. You want a certain result, so you set up the sample, the questions, etc., all to give you the answer you want, then you 'leak' it to the media.

[maseman]

19. You are spot on

The less random and sloppy the "poll" the less accurate and scientific the results will be.

[grantcart]

20. Not true the most effective polling is where an exhaustive sample is taken from

bellweather units.

Whether it is a precinct or a county there are key bellweather precincts and counties that are very effective indicators of changing public sentiment. Its not unusual for campaigns to try out particular mailings on a single area and then go back and recanvass it.

This is also proven in the private sector. When rolling out a new product or brand they will not take random sampling across the country but do an exhaustive roll out in a key demographic areas.

The reason that campaigns can afford to do it is because they will normally use volunteer workers while a private polling operation could never bear the cost for the constant canvassing that the methodology requires.

[RichardRay]

28. Sample design is dependent on random selection

That doesn't mean you have to take the entire country randomly, though. If you have 'bellweather' areas that you believe are highly indicative the the national population then the sample may be restricted to those areas, but then its accuracy is totally dependent on the accuracy of your assesment of that restricted Population, if it's representative, then your result will be useful, if you got it wrong, your result is an accurate representation of an erroroneos assumption.

In essence, you can make two kinds of errors: you can sample badly, and you can sample the wrong population. A funky sampe from a really great population might be more accurate than a better sample from a less representative area, but a better sample of the better area would be better still.

Maybe the exhaustive sampling is sampling really well chosen populations?

[yellowcanine]

9. They aren't. They are just using different likely voter models. They may have more information

about where their ground organization is better or worse, that's all. They are still making assumptions about turnout that cannot be verified before election day.

[Blackhatjack]

11. There is an old joke about what you get when you do 'internal polling'...

... that I cannot tell here, but it comes down to what you make of the results.

[rufus dog]

12. The McLame surrogate I saw

was very evasive regarding the internals, but you could read a lot into his words. Basically he argued that the polls were in the margin of error along with factoring in undecideds. So if Obama has a seven point lead in the state and the margin of error is 4 AND undecideds are at 5%, then the internals favor McCain. They are assuming the polls are off by the full margin of error and the undecideds will break for McLame.

[Drunken Irishman]

14. Internal polls aren't better than real polls.

John Edwards' campaign leaked their internal polls in December last year and it showed Clinton winning Iowa with Obama coming in 3rd.

http://politicalwire.com/archives/2007/12/07/edwards_internal_poll_shows_dead_heat_in_iowa.html

Obviously the internal poll was more generous to Edwards and less generous to Obama.

[aaaaaa5a]

15. I honestly don't think internals are any better than real good independent polls.

Remember the Democratic Primary? Hillary Clinton's internal polls had her blowing out Obama in Indiana and this/close to Obama in North Carolina. And the exact opposite happened on election night.

I think years ago, internal polls may have had an edge. But there are lots of good independent pollsters today. And thanks to the internet, I honestly, don't think internal polls reveal that much more than savy outside political junkies have.

[jobycom]

18. The polls have different purposes.

The national or state polls are designed to measure overall voter preference and sentiment. As such, they are general.

The internal polls are crafted for more specific goals. For instance, if you want to know the best demographic to campaign for in Texas, you can craft an internal poll which tests several of your ideas against those demographics. Hypothetically, say, you can do a poll which shows that your opponent has strong support amongst under 40 white females (or whatever, just a hypothetical), but that his support amongst Hispanic males is weak, and if you campaigned on issues strong to them, you might pry enough votes away to win the state.

So the internals are more specific, and can be crafted to answer more specific questions. Also, since they aren't released to the public, you don't have to worry about hurt feelings. You can poll how many white people really won't vote for a black candidate, and you can poll what issues or phrases you can raise to heighten the racist feelings in similar voters. For instance, you can find that white independents think of African Americans as untrustworthy, or as Affirmative Action employees who are given rather than earning jobs, and so you can work seemingly innocent phrases into your advertisements to play on this. Stuff like "What do we really know about Barack Obama?" Or "He's the greatest celebrity in the world, but what has he really done to earn that status?" You can thus claim you aren't being racist, when you know good and well that you are trying to trigger racist fears that your polls show you should trigger to win.

So you use these internal polls to craft a strategy and to uncover trends that could help you, and then you create polls that take these results into account. National polls might miss some of these trends (especially where the so-called Bradley Effect comes in), or they might craft formulas that assume a certain demographic will vote one way, whereas your own polling shows that that demographic is a false grouping, and that it is really two demographics that you see split. So your formula would be more accurate (you think). Say, for instance, a poll showing that Hispanic males in Texas support Obama. A national poll might leave it at that and assume that their numbers are representative of all Hispanic males, whereas an internal poll may discover that in fact there are two distinct demographics being lumped as "Hispanic Male," and that one of those groups will not vote Obama, so that the poll results on the nationals are off.

Anyway, if McCain is claiming his internal polling shows a win, that's part wishful thinking, and maybe part optimistic polling. He may be seeing that the national formulas are outdated (in this election that wouldn't surprise anything, since there are so many firsts here), and that his internal polling shows that he is more competitive in certain areas.

Also, there's the southern effect. Truman beat Dewey, despite Truman being considered a joke and Dewey a shoe-in, because the South voted as a solid block for the Democrats. The South always votes this way, but it was harder to measure in the national polls in 48. Look at what's happening now. Everyone is focused on gender and race, and no one is considering regional aspects. This is the first election in a generation without a southerner on either ticket. If the Democrats win, it will be the first time since FDR in the 30s that we won without a southerner on the ticket. The South now votes Republican, and the only elections we win are ones where we split the South. So, what will happen in the South? The polls show some splits, but will that occur? The nationals are hesitant to delve too deeply into racist sentiment in the South, but will it cause a surge of McCain votes that the nationals are missing on election day? I think that's more debatable than we assume.

One more thing--not that anyone has read this far. There's no doubt that Obama's supporters are the most fired-up and committed, and McCain supporters are disappointed in him, but will that translate directly to votes? A hesitant vote is still a vote. So, on election day, it's arguably conceivable that a lot of Republicans who are polling as weak McCain supporters, or who are refusing to be polled at all, will walk into the polling booth, and their habits and their old racist beliefs will take over. They may hate Bush and not like McCain much, but they still might turn out in droves to defeat the Democrat, just because that's what they do. Internal polls could pick this up more than national polls, which ask more universal, less specific, questions.

I fully expect Obama to win easily. I'm just pointing out what McCain's numbers could be showing that could give him hope. I suspect his numbers give him some reason to hope, but that his certainty is just campaign posturing.

[FVZA_Colonel]

21. Interesting Analysis.

I fully expect McCain's "weak supporters" to pull together for him on the last day, but Obama will still come out ahead (not just because his "weak supporters" probably will as well.)

[leftist.]

23. Thank you jobycom!

I did read all of it :) Thank you for the well thought out explanation - the Truman paragraph taught me quite a bit!

[aaaaaa5a]

27. Good Post!!! n/t

[kennetha]

30. Excellent and informative post

Wish there were more of this kind of analysis here on DU

Thanks a lot for the work and thought you put into that.

[progressivebydesign]

22. because they are paid for BY the candidates, and are actually used for leaking...

..they're BS, basically. At one time internals were useful, but now they are a tool to try and convince the media that you're doing better than you are.

[sfam]

25. My internal poll says Obama is up by 9 points nationally...

Yep...mine is definitely better!

[endthewar]

26. All you have to do is change your "likely voter" model to get whatever results you want.

Desperation time for the McCain campaign. They're just trying to keep morale up.

[Spike89]

29. Strength of support can be critical

I wouldn't be surprised if McCain's internals show a lot of "leaner" votes in PA. These are not people who are listed as undecided, but rather those that could be swayed by an "October surprise" or even a very nasty attack push the last weekend. Almost all the polls ask how committed your vote is, but it is a) too complicated to easily work into the results for general news, b) a "squishy" and thus unreliable question (how many people would change their vote if their preferred candidate actually killed and ate a baby on live TV, what about if they changed a key position unexpectedly, how many could be swayed by something like another terror attack?).

Polls are great, but even big leads can be "soft" and tiny leads can be very hard set (what would it take for you to vote for McCain?) Internal polls won't probably have the gross numbers much different from the external polls, but they will have lots of the break down numbers that can be huge. If you're behind in a state by 10% but your polls say 50% of the opponents supporters would/might change votes if they thought the opponent would raise their taxes, you'd feel pretty confident that you could still win, if only you could make the case.