Cooking the Polls: Democrat Oversampling Producing Phoney Results

www.seanrobins.com ^ | October 19, 2006 | Sean Robins

[seanrobins]

While it may appear that Republicans generally are getting hammered in a wide variety of polls going into this last month before the November mid-term elections, what is almost never explained by the pollsters, or their media conspirators, is that the polling data is deliberately skewed to favor the Democratic outcome. (Gee, who would have guessed?)

The scam is a simple, yet thoroughly devious one. The polling service simply "samples" a disproportionately large number of Democrats versus Republicans in its poll. Taking as an accurate generalization that Democrats poll one way for the most part, and Republicans another for the most part - such Democrat-oversampled polls result in Democrat-leaning results.

Today's case in point, brought to light courtesy of NewsBusters, is that of the latest NBC News/Wall Street Journal poll, in which 1,006 registered voters were questioned during the October 13-16, 2006 period, concerning a variety of standard political issues. The margin of error reported was +/- 3.1% Some of the "highlights" of the poll - those questions which were the primary object of the poll, included the following:

In general, do you approve or disapprove of the job that George W. Bush is doing as president?
Approve	38
Disapprove	57
Not Sure	5

In general, do you approve or disapprove of the job that George W. Bush is doing in handling the economy?
Approve	44
Disapprove	52
Not Sure	4

In general, do you approve or disapprove of the job that George W. Bush is doing in handling the situation in Iraq?
Approve	33
Disapprove	63
Not Sure	4

What is your preference for the outcome of this year's congressional elections--a Congress controlled by Republicans or a Congress controlled by Democrats?
Republican-Controlled Congress	37
Democrat-Controlled Congress	52
Not Sure	11

Generally speaking, do you think of yourself as a Democrat, a Republican, an independent, or something else? (IF "DEMOCRAT" OR "REPUBLICAN," ASK:) Would you call yourself a strong (Democrat/Republican) or not a very strong (Democrat/Republican)? (IF "INDEPENDENT," ASK:) Do you think of yourself as closer to the Republican Party, closer to the Democratic Party, or do you think of yourself as strictly independent?
Strong Democrat	21
Not very strong Democrat	7
Independent/lean Democrat	15
Strictly Independent	15
Independent/lean Republican	10
Not very strong Republican	9
Strong Republican	18
Other/Not Sure	5

This last question reflects the "heart" of the survey's demographic in terms of political affiliation. Note: Although respondents are initially asked whether they are registered - and only those answering that they are continue - the pollsters do not as specifically about party registration. The question is couched in terms of a respondent's views of him or herself poltically.

The results of this question demonstrates the skew in terms of Democrats vs. Republicans. 80% of respondents described themselves in terms of either being or leaning either Democrat or Republican - 43% as Democrat - 37% as Republican. This represents an oversampling of Democrats by 16 per per-cent. When coupled with the margin of error, this represents a very significant statistical advantage of Democrats over Republicans, due entirely to the composition of the sample as determined by the pollster.

What does this mean? Well, it is not possible to know with certainty what the results would have been had the sample not been so significantly skewed. In other words, we do not know how many Democrats approved or disapproved versus how many Republicans. Nor can we know which way - if the sample were not skewed - how the "extra" Republican opinions would have responded. However, we do know that it would have made a significant difference.

If we assume - reasonably though not absolutely - that the omitted Republicans would have said that they approved of how the President was doing his job, this would represent an 8% shift in the numbers. 38% approval and 57% disapproval suddenly becomes 46% approval and 51% disapproval. A five-point difference which, factoring in the +/- 3.1% margin of error, results in a virtual tie between those who approve and those who disapprove of the job George W. Bush is doing.

Even more significantly is the change in the numbers on approval of how the President is handling the economy. Making the adjustment, a 44% approval and 52% disapproval, becomes a 52% approval and a 44% disapproval. A complete reversal.

As to the handling of Iraq, pre-adjustment it is 33% approval and 63% disapproval. Making the adjustment for Democrat oversampling, the results become, 41% approval and 55% disapproval. This results in an 11% differential when considering the margin of error. Not terrific numbers, but much better than what the pollsters would have you believe. Better numbers than this particular poll has reported in over a year. (Of course, without the sampling demographics for the prior NBC News/WSJ polls, we don't know how badly they were skewed Democrat.)

Finally, let's look at the most weighty question: The preference as to Democrat-controlled versus Republican-controlled Congress. The survey reports this as: 37% preference for a Republican-controlled Congress, and a 52% preference for a Democrat-controlled Congress. Again, the 16% skew in the sampling in favor of Democrats renders these results completely incompetent. Making the adjustment as before, the results change stunningly: 45% voice a preference for a Republican-controlled Congress, and 44% for a Democrat-controlled Congress.

"Fiddling the figures" in this one NBC News/Wall Street Journal poll, simply by oversampling Democrats, leads to vastly skewed, erroneous and utterly unreliable results.

But this is what the mainstream media wants. Remember the old adage: Figures lie, and liars figure. That is precisely what is going on in this poll, and in the dozens of others out these the past few weeks which would portend doom and gloom for the Republicans.

[seanrobins]

I thought that the math involved was pretty simple. Let's try again...

If A=45% and B=55% and the total number of respondents in the sample is, let's say, 200 - to make it easy - then A, which is 45% alse equals 90 individuals and B equals 110 individuals.

With me, so far?

A=45%=90
B=55%=110

B is greater than A by 20 persons
B is also greater than A by 18% NOT 10%

The difference between A and B, and a function of the total (100%) is 10%

But since what I was discussing was the degree to which Democrats were being oversampled by the pollster, the significant number is the % which Democrats exceeded Republicans in the sample.

Percentages of percentages - as you put it - ARE significant, when discussing in releative terms, how great an oversampling was made.

There does not necessarily need to be exactly equivalent numbers of Democrats as Republicans in the sample, but it should approximate (in this national poll) the national breakdown of Democratic vs. Republican registrations.

What HAS been happening, and what is significant, is that most if not all mainstream media driven polls are being "fiddled with" in terms of the sample.

Even you should be able to handle the following math:

If the sample was 100% Democrat, the approval rating of George W. Bush would be.......(can you guess?)

[CountryBumpkin]

Exactly what has been happening in "polls" for the Maryland Governor's race. The "sampling" tends to focus on illegal alien paradises like Montgomery County, ignoring the entire central and western part of the state that is heavily Republican. Governor Ehrlich (R) has done a fantastic job for the past four years (wiped out the state debt, no real tax increases to speak of, booming job market). The libs are desperate to paint him in a bad light.

[bray]

Excellent Post. These Phony polls are to keep the talk of a Rat Landside in the Echo Chamber. After we win...silence!

Pray for W and Our Troops

[Paloma_55]

A Survey I would like to see...

1) A believe the war in Iraq is going a) Well b) Not well

2) Generally, newspaper reports I have read indicate the war in Iraq is going a) Well b) Not well

3) Generally, TV News reports I have read indicate the war in Iraq is going a) Well b) Not well

4) My opinion is shaped mostly by a) TV News b) Newspaper Reports

[sizzlemeister]

Even more dangerous... oversampling of Dems in exit polls and then posting the early 'results'. This happened in the last election. This kind of crap aims to keep Repubicans/Conservatives at home on election day. All the proof required to demonstrate which team the MSM bats for is to look at the 2000 Election. Florida was called for Gore with an hour left to vote in the (overwhelmingly Republican) panhandle.

Absolutely. I heard Rush briefly mention exit polls recently, although admittedly BOTH of my kids have been home all week due to the surprise storm in WNY so I haven't listened as closely as I usually do. I think Levin, Hannity, and Limbaugh, at the very least, ought to start hammering away at the point on exit polls starting now.

[YdontUleaveLibs]

Let me guess you are a Demonrat and actually think there are more Democrats than Republicans in the general population.

[HardStarboard]

>>>If you don't care who is elected, sit this one out. <<<

Correction:
If you want to insure Harry Reid and Nancy Pelosi are formulating the laws you must live by, sit this one out!

[Intolerant in NJ]

This represents an oversampling of Democrats by 16 per per-cent. When coupled with the margin of error, this represents a very significant statistical advantage of Democrats over Republicans, due entirely to the composition of the sample as determined by the pollster...excellent work, but you may have overlooked one important point - to have real predictive value, the sample proportions must match closely the proportions in the underlying population, the "real world". Were this study done in certain congressional districts the actual proportion of rats vs. pubs in the district really might have been 43% to 37%, and in such cases this would have been an appropriate model for trying to estimate the outcome of the election. The good news is that this appears to be a national sample, and to the best of my recollection, the registered number of rats to pubs across the country is about equal, so your complaint of disproportionate sampling in this case would be valid. Moreover, a lot of polls I've seen this year seem to be ignoring this issue of matching the sample to the group being sampled, as well as resorting to other scams such as using "registered" voters (as did this poll) or determing "likely voters" through questionable methods such as self-report (most people know the right thing to say if asked is that they intend to vote, but many won't eventually go to do the deed). It's been forty-plus years since I had that last statistics course, but the lesson of making sure your sample represents your population before you try to make inferences from it sticks to this day.....

[CyberAnt]

I do think we have figured out this is war - but the repubs in congress still think it's a "good ol' boy" network. But that disappeared years ago .. ever since the Clinton came to town.

Actually, Bubba Clinton's legacy is: TOTAL DESTRUCTION OF THE DEMOCRAT PARTY. Poor dems .. when they laid down with that dog, they should known they'd wake up with flees!

[HANG THE EXPENSE]

Hey man, dont waste your time with this guy. Seems to be a little off.

[Salvation]

Now for the real polls.................LOl!

I forgot, we aren't supposed to pay any attention to them because they're all biased.

[CyberAnt]

Me too. The dims are believing their own polls. They did this same thing in 2004 - when Carville put the garbage can on his head. LOL!

I can't wait to see the dems' faces when it all falls apart.

At least that's what I'm believing will happen.

[1035rep]

LOL. The libs are so pumped up with the fake polls....I think we may lose a few this time around.

Then we get the added bonus of Rangel resigning as long as he keeps his word.  I can hardly wait.

[eureka!]

Surprised you were not on this thread. Hence, ping...

[WashingtonSource]

Once upon a time polls had credibility. They were "scientific." Pollsters have completely undermined their credibility because their "polls" are not remotely scientific. They are scams!

[t2buckeye]

I believe there are two main reasons for over-sampling the Dems
1) Naturally, it improves their fund-raising by encouraging their base and gives the media a great way to cheer on their leader
2) It is an effort to use any Dem loss as a way to cry "voter fraud" should the Republican win.
Look at the recent Robert Kennedy book...his whole premise is that polls/exit polls suggested a Dem win and the Reps won in OHio....of course, we know that exit polls can be totally wrong (especially when pollsters are young and obviously of the Dem persuasion as is what happend in 2004...respondents didn't want to get in an argument so they lied)

[NittanyLion]

Finally, let's look at the most weighty question: The preference as to Democrat-controlled versus Republican-controlled Congress. The survey reports this as: 37% preference for a Republican-controlled Congress, and a 52% preference for a Democrat-controlled Congress. Again, the 16% skew in the sampling in favor of Democrats renders these results completely incompetent. Making the adjustment as before, the results change stunningly: 45% voice a preference for a Republican-controlled Congress, and 44% for a Democrat-controlled Congress.

Your numbers don't work out. I ran the numbers assuming 85% of each party prefer that party in power, 5% are unsure, and 0% prefer the other party (historically this is accurate). Under those assumptions it's evident that self-identified independents prefer a Democrat-controlled Congress by a 2-1 ratio.

Now, if we assume the actual electorate will be 35% GOP / 35% DEM / 30% IND (again, historically that's accurate), then the adjusted results would be as follows:

Preference for GOP-controlled Congress: 38%
Preference for DEM-controlled Congress: 47%
Not Sure: 15%

As others have stated, you can't take percentages of percentages and expect it to work out. You have to convert the percentage into real terms, adjust the numbers, then reconvert it back into percentage format. Sorry.

[NittanyLion]

Oops - "and 0% prefer the other party" should be "10% prefer the other party".

[WOSG]

"1,006 registered voters were questioned during the October 13-16, 2006"

Registered voter poll means almost nothing when it comes to elections. Needs to be 'likely voters' to mean much. These indeed are typically Democrat-biased, especially in off-year elections.

Democrats are running anti-incumbent races, but that doesnt

That explains why the folks like Sean Hannity are pumping up the GOP base ... well, they would do it anyway, if we were having a great year, we'd have the cheerleaders pushing for turnout so we could have a filibuster-proof majority.