« Quinnipiac Poll: Kerry Leads in PA by 6 | Main | Paul Krugman Weighs in on the Gallup Problem »

About those Mason-Dixon Polls

By Alan Abramowitz

The Mason-Dixon polling organization has released a raft of polls in key battleground states in the past two weeks. Mason-Dixon is a reputable organization, but if you’ve been getting the feeling that their results this year are a bit skewed, you’re correct. I compared the results of Mason-Dixon’s October polls with the average of all other polls released in October in the same states. So far this month, Mason-Dixon has released 16 polls in 14 battleground states. Here are the comparisons. The number shown is the difference between the percentage of likely voters supporting George Bush and the percentage supporting John Kerry. A negative number means that Bush was trailing. The number in parentheses is the number of October polls in the state.

AverageBush
of AllMason-Dixon
StateMason-DixonOther Polls Advantage
 
Colorado+7.5(2)+4.5 (6)+3.0
Florida+3.5 (2)+0.3(10)+3.2
Iowa+6.0(1)+0.8(7)+5.2
Michigan-1.0(1)-4.2(5)+3.2
Minnesota+2.0(1)-3.5(4)+5.5
Missouri+5.0(1)+4.7(3)+0.3
Nevada+10.0(1)+5.5(2)+4.5
New Hampshire+3.0(1)-2.0(8)+5.0
New Mexico+5.0(1)-2.0(4)+7.0
Ohio+1.0(1)-1.3(11)+2.3
Oregon-1.0(1)-4.4(5)+3.4
Pennsylvania-1.0(1)-3.8(9)+2.8
West Virginia+5.0(1)+2.0(1)+3.0
Wisconsin 0.0(1)-0.2(1)+0.2
 
14 State Average+3.2;(16)-0.3(76)+3.5

In all 14 states, Mason-Dixon’s polls were more favorable to George Bush than the average of all other polls. Based on Mason-Dixon’s polls, Bush was leading in 10 states, Kerry in only 3. Based on the average of all other polls, Bush was only leading in 6 states, Kerry in 8. Given the number of states and the number of polls included in this analysis, this difference is quite striking. In some states, such as Missouri and Wisconsin, the difference between Mason-Dixon and the other polls was trivial. However, in the crucial battleground states of Iowa, New Hampshire, and New Mexico, the difference was large enough to give a rather different impression of the state of the presidential race. Moreover, since the Mason-Dixon polls are often sponsored by prominent media outlets, it is likely that their results have been more widely publicized than those of most of the other state polls.

I have no explanation for why Mason-Dixon’s polls in the battleground states have been consistently more favorable toward George Bush than other polls in the same states. Perhaps the election results on November 2nd will prove that they were more accurate than their competitors. However, it is worth noting that the average of all of the other state polls, a dead heat between George Bush and John Kerry, is closer to what most of the national polls have been showing recently in the battleground states.

Comments

A realclearpolitics (or whatever that poll-compilation site is called) founder was on C-SPAN this morning and Brian Lamb asked him which poll he trusted the most. The guy refused to name a national polling outfit, but with regards to the state-by-state polling, he singled out Mason-Dixon as being the best. So I read with interest what you had to say about it. (The guy also admitted to being center-right in his politics, so perhaps he just likes the pro-Bush results?)

Anyways, I was wondering what Mason-Dixon's track record was in the 2000 election. Like you wrote, maybe M-D is on the mark, and the other polls are wrong.

So why is it that pollsters are not required to offer possible explanations for why their results are out of whack with others? Wouldn't that make for a much more interesting story in the first place?

This is interesting, in that Mason-Dixon has a consistent skew relative to other polls, but I wouldn't get too excited about it. I don't think they're cooking the books by any means, after all, they did correctly predict the winner of all the swing states in 2000.

I noticed this too when looking at them at

http://2.004k.com/latest/

Has any looked at the internals? They might be over-sampling Republicans like Gallup.

On a related topic, the values in the futures markets at tradesports.com look very skewed toward Bush. In fact, it's so striking that I would hardly be surprised if Republicans were manipulating the market. Perhaps, we should separate them from their money?

I assume Mason-Dixon is "likely" voters. If so, does anyone know their method of selecting likely voters from the pool? It seems they could be falling prey to the same problems that seem to bug Gallup this year, in miscalculating just who may vote this time around. I've always agreed with Ruy that a large turnout and new registrants will favor Dems and that seems to be what is shaping up.

I also read somewhere that Zogy said in an interview that if turnout exceeds 110 mil (vs. 105 mil in 2000) that all bets were off and it was likely the Dems would be in a much better position than it currently seems.

Go Red Sox.

I believe mason Dixon was the most accurate polling outfit in the last election of 2002. having said that, I think all polls have no idea what the turnout model is going to look like this go around. Hopefully we will get a very high turnout.

MSNBC had a headlines

Polls: Kerry under pressure in 'blues'
New polls keep Bush 'in the red'

Kerry is behind in OH, tied in MI PA. Guess which polling service they used ;). They did not reference other polls, just Mason-Dixon.
MI is the worse, every poll has him ahead and the gap is greated than MOE in most.


http://www.msnbc.msn.com/id/6289208/

http://www.msnbc.msn.com/id/6301044/
Meanwhile in Michigan, a state Gore won by more than 200,000 votes, Kerry and Bush are in a statistical tie, with Kerry at 47 percent, Bush at 46 percent and six percent undecided.

When Mason-Dixon did its previous round of polling for MSNBC and Knight Ridder newspapers in Michigan in mid-September, Kerry stood at 47 percent, with Bush at 41 percent and 10 percent undecided. Bush has thus gained five points in about a month.

I've always thought of Mason-Dixon as a Republican-leaning poll. It doesn't have that reputation among professionals? It sure seems to me that they've been reporting results in 2000, 2002, and 2004 which were skewed Republican.

According to Joe Monahan's report (http://joemonahansnewmexico.blogspot.com/) the Mason-Dixon poll in New Mexico under represents independents. If this is true in the other battle ground states, it may account for the general discrepancy between MD and other pollsters.

The new Times Poll as just reported by CNN also shows Bush above 50% with both RV and LV and Kerry trailing by about 5-6 pts. So now there are atleast 4 major polls showing Bush at 50 or above (Post/ABC, Newsweek, Gallup, Time). Is the Mason-Dixon poll reflecting the trend towards Bush extending to battleground states ? That's an eerie possiblity...

Without any internal information about M-D's polling methods, it's not possible to dissect why their results are so skewed from those of other sources. However, I don't think it's merely wishful thinking to assume that there is some kind of systemic bias in their approach. If there were not, one might find variability from other polls, but not such consistent, one-directional variability. The problem could arise from their sampling method, their questions, their interviewers or other techniques, or whatever weighting assumptions they employ. Whatever it is, it is actually sort of "encouraging" to see the consistency in their bias, relative to other polls, because the more often it occurs, the more indisputable the conclusion that such a bias exists, systematically, in all Mason Dixon results. This would suggest, for example, that a 3-pt Bush lead in any particular M-D poll is indicative of approximately a tie in that state.

So, we should not be surprised that this very afternoon, the Mason-Dixon Polling from Iowa, Minnesota, Pennsylvania, Ohio and Michigan were the only data featured during CNN's Inside Politics!

Mason-Dixon is simply projecting that 3% of Democrat votes will not be counted.

Looking at the Iowa poll, is it true that 59% of Iowa voters are 50+ years old?

*smirk*

I don't think that's true. A quick Google look at population tables suggests this poll is skewed towards older people. Mason-Dixon seem to be running into the problem of "Who's answering the phones?"

As long as you're looking at specific polling companies, can you explain how Survey USA generated almost the same exact poll numbers for Bush and Kerry between the 18th and 20th in NV, CO, IA, MO and AR? How can all five states suddenly have Bush at 51% (IA, MO, AR) or 52% (NV, CO), while Kerry is at 45% or 46% (AR)?

assuming the mason-dixon polls are correct.
If the undecides breaking for the challenger rule stays true, and dem's due a better job at turn out then 4 years ago, the MD polls are'nt all that bad.

46-45 bush in Ohio.- Pretty good position for kerry to be in, with that many undecides. it's hard to see bush winning enough undecides to hold on to this lead.

MN. Bush 47-45.- Def. Worrying, but for an incumbent pres 47 percent is'nt all that good this late in the game.

Wisconsin.-45-45. With that many undecides left, it's hard to assume Bush will win.

If kerry wins these 3 states, he will win the election even if he lose NH, FL, and NM.(Mason Dixon shows bush leading in NH & FL by 3, and NM by 5).

NH and FL are still close enough in the MD poll that strong turn out of the Dem base, or a higher % of young voters, combined with Kerry doing well with the Undecides show kerry still has a good shot of winning the election.

The truth of the matter is that M-D had extremely accurate results in the 2002 midterms. They were by far the best. Way better than Zogby or SurveyUsa. I'm not sure how they did in 2000 but I sure hope they are off this year. I must admit there polls scare me a bit given the incredible job they did in 2002.

The Knight-Ridder articles reporting the Mason-Dixon polls state:

"Those interviewed were selected by the random variation of the last four digits of telephone numbers. A cross-section of exchanges was utilized in order to ensure an accurate reflection of the state. Quotas were assigned to reflect voter turn-out by county."

This appears to say that the poll is not weighted demographically, except to match a geographical distribution. Such an approach would lead to a significant undercount of minorities and under-24 voters, thus skewing the results toward Bush.

It's hard to take seriously any poll which would call itself the Mason-Dixon Poll.

Was "Johnny Reb Polls" already taken?

My fear is that the fraudualent/skewed polls are being used by the corporate mediawhores in conjunction with thier own slanted mis-reportage to keep the election close enough to steal.

Close enough in our minds' eye that the various vote fraud/ disenfranchisment/ suppression schemes being employed won't be widely questioned by the public because the impression that it's so close, so tight, or that Bush might be ahead are now solidly implanted in the mass consciousness.

This psyop/ psy-war (constant re-enforcement of Smirky as maximum leader) has been ongoing since his installation four yrs. ago, but especially since 911...

Mason-Dixon has been polling statewide and regionally in North Carolina, as well, producing numbers that are consistently more pro-Bush by a few points than other state polls. (The latest was 51/43, where Zogby and others give Kerry around 47-8.)

And Mason-Dixon the Gannett pollster of choice? I ask because its polls are appearing in the Asheville Citizen-Times...

I think we will find that M-D's polling results will be pretty close come election day. Hopefully they will tighten up this week and the "undecideds" factor and the "unexpected voters" factor will tip the scales for Kerry. Here's a review of how they won the 2002 midterm polling hands down:

NATIONAL COUNCIL ON PUBLIC POLLS POLLING REVIEW BOARD

February 05, 2004

FOR IMMEDIATE RELEASE: THURSDAY, DECEMBER 19, 2002 NATIONAL COUNCIL ON PUBLIC POLLS POLLING REVIEW BOARD ANALYSIS OF THE 2002 ELECTION

Most organizations conducted polls in only one or two states. They accounted for 95 of the 159 polls. Five organizations conducted 64 polls. These organizations worked in three or more states. Mason-Dixon did the greatest number. They conducted 23 polls in 16 states. Only one of their polls had the wrong candidate winning. Zogby International did 17 polls in 12 states, and had 5 incorrect winners. Zogby also conducted partisan polls, which are not included here even though they were publicly available. Research 2000 did 13 polls in 9 states, and had 2 wrong winners. The Gallup Organization did 7 polls in 4 states with 1 mistake. And Quinnipiac College in Connecticut conducted 4 polls in as many states. None had the wrong candidate winning. The 95 polls from all other organizations had the wrong candidate winning in 13 of the contests. In only one of the 22 races where the wrong winner was named was the mistake due to a small lead in the poll for the losing candidate when the other candidate won a very close race.

Skewed? Yes, indeed, the Mason-Dixon polls appear to be skewed in favor of reality. Remember that next time when you're pumping Gallup and Zogby.