Media bias/statistical question

[lendaddy]

Media bias/statistical question.
his is the data of those polled in the recent CBS poll showing Bush's approval at 34%. And I think the margin of error is listed at 3-4% or atleast something like that.



Now assume the only poll question was:

Are you a Republican or Democrat or Independent?

The poll would then conclude that:

26% of Americans are Republicans

40% of Americans are Democrats

33% of Americans are Independents

With a margin of error of 3-4%

Now this is obviously waaaaaaaay off. So, one of you statistics guys please explain how (sans intentional shaping) the scientific method broke down so badly?

[RallyJon]

Most Republicans are *******s and/or have better things to do than take a CBS poll and hang up on the pollster?

[TheMentat]

I'm not sure I understand why it is waaaaaay off... (I think I need you to point out the obvious for me ;-)

[RallyJon]

I think you also have a large number of Republicans who don't want to be painted with the media's broad brush that Republican = GWB. I think some Republicans have found that going around shouting, "I'm a Republican" these days tends to associate you more with the current administration than with traditional Republican values.

Face it, Republicans tend to be more conservative in many ways, and that can mean less willing to publicly divulge information. I, for one, know that it's much more fun to mess with pollsters by saying I'm a Democrat.

Look at the polls from just before the last election. Look at the exit polls that got it so wrong.

Bottom line is, current political polls using current methodology are NOT accurate, not plus or minus 3.4%, not at friggin' ALL.

[TheMentat]

What is wrong with the poll numbers?

[lendaddy]

Originally posted by TheMentat
What is wrong with the poll numbers?

You think 40% of Americans are Democrats and 26% are Republicans?

[TheMentat]

Originally posted by lendaddy
You think 40% of Americans are Democrats and 26% are Republicans?

Ahhh.... well I'm not sure, but it does seem low...I think RallyJon's explanation sounds reasonable. To be fair, there's nothing wrong with the actual results... the problem seems to be in their interpretation of the results. A more correct interpretation would be:

"Based on the the polling of 1018 respondents, we estimate that 26% of Americans would declare themselves to be Republicans +/- x%" (as per RallyJon's explanation)

I'm not sure what the sampling error would be on 1018 resondents though, but 3% does not seem unreasonable... The sampling error is based on a number of factors:

# of responses

The actual distribution of of the underlying population (50-50 vs 80-20)

# of potential answers (I think)

I cant seem to find the right section in all of my old texts, but Wikipeda has a small blurb on it:

http://en.wikipedia.org/wiki/Opinion_poll

They also could have effe'd up the entire sampling process by only phoning people in California for example (but I'd imagine they control for things like geographical heterogeneity).

[Superman]

I've done some statistics though I am not a mathematician. One day while waiting to testify in a hearing I heard an attorney ask a statistician, who happened to also be my prof at the time, "Dr. Ott, in your opinion, do statistics lie?" Dr. Ott looked at the ceiling as though he had never considered this question before then carefully lowered his gaze and made the perfect response. "No."

You see, numbers are our slaves. The statistics themselves are 100% accurate in a sense. The to major problem sources are 1) whether your data is clean and 2) whether your interpretation of the results is faithful. Estimating proportions is a matter of sample size (of course), degree of confidence you target and the population's characteristics (small proportions are more difficult to quantify/estimate).

In the instant case, it may be that the respondents' perception of the question(s) they were asked left freedom to reject both parties. It looks like it. I infer that respondents felt free to reject either or both parties in favor of "Independent." Had this option not been available, who knows how many of those "Independents" would have chosen a "D" or and "R" response.

And quite frankly, I am not surprised that the remaining "D' responses outnumbered the "R" responses. For perhaps several reasons:

1) "R" people consider themselves very independent and that has resulted in some trafficking between stated parties. That is, the Republican Party has managed (to their credit) to capture many independents in recent elections. There seems to be a rather fickle group out there that the Republican Party has called its own recently that do not actually belong to them in a permanent sense.

2) The Republican Party is the easiest party to hate today. For at least two reasons. First, be careful what you wish for. Your party is in power now, and that carries responsibility and blame. Whether it is their fault or not, Republicans are currently responsible for running the show. And the show is not going well....at least not in the minds of folks who are open to the facts and who are independent thinkers.

3) Whether you like it or not, this new development where the Republican Party is in control is just that....new. Until recently, this nation had more stated Democrats than Republicans. When that happens again (soon, I suspect) Mul's debasement of Democrats will be treachery. We will see his lack of patriotism, I suspect.

[RoninLB]

Rep politicians in general have lost the reasoning that put them in power. If they don't straighten their act out elections might as well be decided by Quaalude poker.

[Mulhollanddose]

Originally posted by lendaddy
You think 40% of Americans are Democrats and 26% are Republicans?

Including the felon, dead and illegal immigrant vote, tried and true Democrats...yes.

[RoninLB]

This whole statistical poll bs is a White House plot to trash the Dem party.

The gamblers are the $ supporters. Statistically who gets a better return on their bucks? Some people have to learn the hard way and a consistent winner sooner or later draws previous losers. I'm betting that CNN believers will soon say "How did that Bush crook succeed again? How did the RNC pull off a winning election again?"