Skewed Sample Distorts Kos GOP Poll

Markos Moulitsas of the Daily Kos has commissioned a poll that seeks to gauge the political views and socio-religious beliefs of the GOP base. Some of the results are quite simply stunning. Thirty-nine percent of Republicans want Barack Obama impeached, 36 percent do not believe that President Obama was born in the United States, 63 percent believe that he is a socialist, 21 percent believe that ACORN stole the 2008 Presidential election, 31 percent believe that Barack Obama is a racist who hates white people, 23 percent want their state to secede from the Union, 31 percent want all contraceptives banned, only 8 percent believe openly gay men and women should be allowed to teach in public schools, 77 percent believe in a literal interpretation of the Bible.

It is curious that of the nearly two score of blog stories on the Daily Kos/Research 2000 poll, not one looked at the sample, though to be fair to Ezra Klein, he at least did have the thought that "maybe just maybe the sample might off" cross his mind. A poll is only as good as its sample and this poll oversamples older (37.09 percent of the sample is over the age of 60), southern (42.24 percent of the sample hails from the old Confederacy plus Kentucky) men (56.16 percent of the sample are men). It is a great poll if we wanted to get insight into the views old southern men who vote Republican. While that's certainly the stereotype, the face of the GOP is broader than that.

Last May, the Gallup Organization did find that the Republican base was heavily white, conservative and religious based on a poll of 26,314 national adults, aged 18 and older, a sample size more than ten times larger than the Daily Kos/Research 2000 poll.

More than 6 in 10 Republicans today are white conservatives, while most of the rest are whites with other ideological leanings; only 11% of Republicans are Hispanics, or are blacks or members of other races. By contrast, only 12% of Democrats are white conservatives, while about half are white moderates or liberals and a third are nonwhite.

The results show clearly that the Republican Party today is first and foremost a political entity dominated by white Americans. Eighty-nine percent of rank-and-file Republicans are non-Hispanic whites, leaving just 5% who are Hispanic (of any race), 2% who are black, and 4% of other races.

Further, by well over a 2-to-1 ratio, whites who identify as Republicans claim a conservative, rather than a moderate or liberal, ideology (or have no opinion when asked about their ideology).

The ethnic mix the Daily Kos/Research 2000 did get right. Their sample is divided between 89.21 percent white and 10.78 percent non-white. This is certainly a problem for the GOP. While the rest of the country is increasingly multi-racial, the GOP remains overwhelmingly white. Just under four percent of Republicans are African-Americans. A little more five percent of the GOP is Hispanic but about half of those are Cuban-Americans. And only two percent are Asian.

But the poll oversamples men by about four percentage points. While the GOP does face an overall gender gap, the composition within its own party is fairly even split. It is independents that account for the gender gap when election time rolls around. As Gallup noted in May 2009:

Women's affinity for the Democratic Party looks even stronger when independents' partisan leanings are taken into account. By this measure of party identification, Democrats currently enjoy a 22-point advantage over Republicans, with 57% of women identifying as Democrats or saying they are independent but leaning Democratic, compared with 35% who identify with or lean to the Republican Party.

But where the poll is really off the mark is in its age breakdown and its regional distribution. The Institute of Southern Studies correctly I think points to the distortion that this bias creates:

The poll has one big flaw: 42% of those polled came from Southern states -- way out of proportion with their share of Republican voters nationally.

This over-sampling of Southern Republicans (846 total) skews the national results, but it also means the data is especially rich in giving us a picture of the views held by GOP voters in the South.

And the picture is unmistakable: On almost every issue, Southern Republicans are far to the right of their national GOP brethren. In fact, GOP Southerners appear to be the driving base for some of the most extreme views circulating in the Republican Party today.

To measure this, normally we'd compare the Southern results to the national average and see what the difference is. But since the poll disproportionately surveyed Southerners to start with, instead I looked at how the Southern answers compared to the next most conservative region.

For example, here are four questions the poll asked Republicans about President Obama, with the Southern poll numbers compared to the next-highest region (in each of these cases, the Midwest):

QUESTION: Should Barack Obama be impeached, or not?
South: 42% yes
Next-highest region: 38% yes
Southern difference: +4%

QUESTION: Do you believe Barack Obama was born in the United States, or not?
South: 43% no
Next-highest region: 33% no
Southern difference: +10%

QUESTION: Do you think Barack Obama is a socialist?
South: 67% yes
Next-highest region: 61% yes
Southern difference: +6%

QUESTION: Do you believe Barack Obama wants the terrorists to win?
South: 28% yes
Next-highest region: 22% yes
Southern difference: +6%

Here are the demographics used in the sample versus a more accurate picture of the composition of the GOP as compiled from various sources.

Demographic Sample Size Sample Pct. Actual
Men 1125 56.16% 52%
Women 878 43.83% 48%
White 1787 89.21% 89%
Other 216 10.78% 11%
18-29 178 8.88% 15%
30-44 418 20.86% 26%
45-59 664 33.15% 37%
60+ 743 37.09% 22%
Northeast 217 10.83% 18.60%
South 846 42.24% 36.32%
Midwest 437 21.82% 25.45%
West 503 25.11% 19.63%
Source: Daily Kos/Research 2000,
Gallup, Pew Research Trust, CNN

In short, the Daily Kos poll has a bias that oversamples Southerners who are more extreme in their views (the poll also drastically undersamples the generally more moderate Republicans in the Northeast by over eight percentage points) and thus paints the GOP as more extreme than they really are. The Daily Kos poll is an inaccurate reflection of the national GOP but likely an accurate picture of the views of its Southern base which nonetheless does account for over one third of its electoral strength nationally. And that Southern base plagues the national GOP to a degree that cannot be overstated.




Tags: Republican Party, Demographics, Daily Kos (all tags)



What is there to say.

We are a fucked up country of quite diverse people, people who routinely vote against their best interests in order to expell their anger. Anger about what? Got me.

by MainStreet 2010-02-03 03:04PM | 0 recs

at whatever the pretty faces on TV tell them to be angry about

by ND22 2010-02-03 11:20PM | 0 recs
Your diary is missing somethinng.

Have you reached out to KOS or Research2000 for an answer as to why they sqewed the Poll? Wouldn't it have been proper form to ask the object of your criticism to comment?   Just asking.

by eddieb 2010-02-03 03:20PM | 0 recs
Skewed? Not so much

Perhaps the demographics of the poll are skewed, but if you look at the cross tabs it is still representational of the Republican mindset.

Look at the numbers - take the overall opinion (which you claim is what is skewed, and then the numbers from the most moderate - either the youngest, western or northeastern - elements and you get relatively the same picture:

Question:  Do you think Barack Obama is a socialist?

overall -    yes - 63  no 21-  not sure- 16       "most moderate" -`yes - 57  no - 25  not sure - 18

Question:  Do you believe Barack Obama was born in the United States, or not?

overall -   yes - 36  no - 42 not sure - 22        "most moderate" -yes - 29 no - 47 not sure - 24

Question:  Should Barack Obama be impeached, or not?

overall -  yes - 39 no - 32  not sure- 29           "most moderate" - yes - 34 no - 35 not sure - 31

Question: Do you believe Obama is a racist who hates white people?

overall -   yes - 31 no - 36 not sure - 33         "most moderate" -  yes - 27 no - 40 not sure - 33

Question: Should public school students be taught that the Book of Genesis in the Bible explains how God created the world?

overall -   yes - 77 no - 15 not sure - 8         "most moderate" -  yes - 70 no - 23 not sure - 7

 And on and on....

Even the most "moderate" of republicans are predominantly crazy.


by mjshep 2010-02-03 03:36PM | 2 recs
Weak and superficial

You either ignored or glossed over that the sample was of self-identified Republicans rather than registered Republicans.  Large swaths of right-leaning voters no longer want to identify themselves as Republicans, even if they still register and vote that way.  This phenomenon is much more likely to exist where 1) The typical Republican doesn't hold such extreme views, and 2) Where they are not in such a small minority.  In other words, outside the South.

So, it's perfectly normal for the South to be overrepresented in this sample compared to national registration averages.  Whether this accurately reflects the 2010 breakdown of voters remains to be seen and will be brought up in the likely voter model debates.  But do you think Research 2000 intentionally overcalled Southerners in order to make their results more sensational?

This is not the first time I've read a front pager here trashing a dKos/R2000 poll for specious reasons.  Jerome called R2000 a bad name (Zogby) for oversampling Hispanics by a point or two in 2008.  Geez, maybe the blog that spawned dKos is getting a little jealous of its offspring's success.

by corph 2010-02-03 04:03PM | 2 recs

This sounds like a real issue, and I'm glad we have the intellectual honesty to focus on it, but bear in mind that we are talking about roughly a 6% overrepresentation of a group that is about 6% more conservative than the median Republican.  So overall, the effect on the bottom-line numbers is not significant.  Instead of 43% of Republicans thinking Obama was born outside the US, maybe the real number is only 41-42%.

The point above about self-identification also sounds reasonable.

by Steve M 2010-02-03 05:20PM | 0 recs
so what

A 4-5% difference is obviously normal for a poll, there should be some groups that are due to statistical sampling. That in turn makes no difference, since a question in which southerners are even 10% more likely to say yes means we're talking about a 1% skew.

That's why the author of the post doesn't compute the results if he forces the geographic fractions to his preferred values, it wouldn not make any meaningfulv difference.

by John DE 2010-02-03 05:24PM | 0 recs
Sort of so what

  At a certain point if Republican membership in a region is too small, it becomes irrelevant what they believe, so maybe giving extra credit to the South is appropriate.  Of course, Brown's election in Massachusetts means one can't be too dismissive, but overall I'd think the opinions of a Republican from a red state are more important than those of a Republican from a blue state.

by whomever1 2010-02-03 06:19PM | 0 recs
between-group differences

Look at the differences between age groups, gender groups, and regional groups -- they are small and probably not significant. So the topline results may be off (but to a certain extent, that's what the MOE is for), but even if we took the most generous subgroups as the actual representation of the GOP it still makes them look completely bonkers.

by aaronetc 2010-02-03 07:33PM | 0 recs
Where's your evidence?

You knock the age distribution, but you have no evidence that the percentage of self-identified Republicans is any higher. Same with the geographic numbers. 

This poll wasn't weighed, yet calls were made random dialing across the US. Turns out that southerners and older folks were more likely to call themselves Republican. Is that really such a surprising result to everyone? The GOP is increasingly older and more southern. That's a fact. That the polling bears that out isn't surprising, it's a reflection of the current trend.

No polling is perfect, and the numbers might be flubbed, but you certainly haven't provided any evidence to that effect.

by kos 2010-02-03 07:51PM | 3 recs
The sample doesn't distort the poll

It's just something that people should understand about it and bear in mind when drawing conclusions from it. 

by JJE 2010-02-04 01:36AM | 0 recs
I would still find it frightening

if even 10 to 20 percent of Republicans say things like Obama should be impeached and my state should secede from the U.S.

I was surprised southerners weren't more different from other Republican respondents, frankly.

by desmoinesdem 2010-02-04 06:24AM | 1 recs
Even If The Poll

only sampled the South, that would be a valid indicator of that part of the nation that has the Senate in a stranglehold via the filibuster.

by Bob H 2010-02-04 03:40PM | 0 recs


Advertise Blogads