Could DU do its own polling?

[cheezus]

Given our number of users and their geographic diversity could DU do polling? Have users sign up to call N random people in their phone book?

We keep seeing so many polls around here, and everyone keeps freaking out. It would be nice if we could do our own large sample survey and have access to the raw data.

Anyone with the knowledge out there? Is this feasible?

[travisleit01]

1. Would it really be random?

It seems like many people might have more contacts that are left-leaning and it wouldn't be a truly random sample.

[cheezus]

2. it'd have to be random

flip open the phonebook, pick a number with eyes closed

[bschoech]

4. On simple random samples

Even that wouldn't be completely random, because a person is far more likely to open a phone book to a central page than to one on the outside or inside. In polling, "random" means that each subject in a particular population must have an equal chance of being selected.

To claim that the poll represents the entire nation, every household in the country would have to have an equal chance of being called. Usually, I think, this is achieved by randomly selecting a series of digits, and using that series as a phone number to call.

If I remember correctly, most telephone polls have a sample size of around 1500, so about 6000 people would have to be called. And then a margin of error would have to be calculated, et cetera.

So conducting an accurate nation-wide poll is a pretty significant undertaking.

[arcos]

3. use a statewide phone book...

and just call random people...

[RoyGBiv]

5. Sorry, but no

Polling is not a simple matter of opening a phone book and calling people to ask them questions. The results you get from something like that wouldn't be a great deal more accurate than a web-based poll. Accurate polling is an involved process, and it only begins with the actual phone calls, which is hard enough in itself. Raw data doesn't mean a great deal. That data must be interpreted, and that's were statistical analysis comes into play. Accurate polling can also be quite expensive unless certain professionals with access to various statistical tools are willing to donate large amounts of their time to such an effort.

I'm not trying to be a wet blanket, and in truth it probably could be done with the vast resource of talents that are on DU. But this is something that takes a lot of time, effort, and expertise. Since there is no basic structure in place from which to start, we're talking about a process that would have to begin from the ground and work up. If the administrators of DU or a group of individuals with enough desire would like to get involved in that kind of organizational process, it could be done, but they'd probably have to quite their jobs to do it effectively. I've interpreted raw polling data for research projects. That's time consuming enough. I don't know that I'd have the patience to start from scratch.

There are people who make their livings from being pollsters, and this isn't because it's an easy job.

[LoZoccolo]

6. We started last week.

93% Kerry, 7% Bush. See now you know CNN is fucking with the polls!

http://www.democraticunderground.com/discuss/duboard.php?az=show_topic&forum=132&topic_id=503445

[TruthIsAll]

7. Here is one way to do it.

Pick a county which is a representative sample of the U.S. and went evenly for Bush and Gore (50/50).

Or, do the same for a county which went for Bush, say 55-45.

Have 100 DUers each make 3 random phone calls in the county.

Ask one question:
If you could vote today, who would it be: Kerry or Bush

Compare the results. A shift of 3-5% from 2000 would be a clue as to how voters are leaning..