Some have wondered as to how well the polling was in 2004 in terms of Bush v. Kerry, indeed, there has been some discussion over whether Kerry underperformed his poll numbers in 2004 and therefore whether that will be the case for Obama in 2008.
For those interested in the question, Charles Franklin provides the following comparison:
Basically if the trend is above the X-axis, the polls measured a Democratic lead and the X’s show the final tally. The 2000 polling trend-line for Gore missed the mark by a about 2.5% and the 2004 number missed by roughly 1.5%–that is to say, MOE level variations. As such, the polling in the last two cycles has actually been pretty accurate.
If people think that they remember the polls being more favorable to Kerry than they ended up to be, it is likely that what is being remembered is that early exit polls in 2004 seemed to indicate a Kerry win, but that those numbers ended up being flawed (for a trip down memory lane see here, here, here and here).
I noted this particular graph via Chris Bowers’ post on Three Big Myths About The 2004 and 2008 Elections.
At any rate, the likelihood of the polls being off significantly in regards to being a predictor (not that that is actually what they are for–they measure opinion at they time they are taken and do not purport to see into the future) is low. However, if they are off, the likelihood is that the deviations from the final polling will actually take the form of a larger Obama victory than the trends indicate because the traditional likely voter models were wrong about who they thought those voters would be. There is, of course, the potential for the Bradley Effect (i.e., white voters telling pollsters that they will vote for a black candidate, but ultimately being unwilling to do so). However, based on what I have read about the issue, the actually Bradley Effect itself in the California Governor’s race in question was not as dramatic as is made to be in the press. Further, I see no reason, except speculation, to foresee such an effect this year.
Sphere: Related ContentThe views expressed in the comments are the sole responsibility of the person leaving those comments. They do not reflect the opinion of the author of PoliBlog, nor have they been vetted by the author.
November 3rd, 2008 at 11:43 am
Obviously the 2000 poll average missed the actual popular vote plurality. But given that Gore won by only about 0.5%, they can probably be forgiven that.
Indeed, the final poll average was almost spot on in 2004, which dispels a myth held by many Democrats (though it says nothing about the exit polls, which are another matter).
On the Bradley effect, independent of the actual Bradley contest, most of the evidence is that whatever effect there has been over the years has dissipated. Moreover, Obama tended to overperform his polls in the primaries much more often than underperform.
November 3rd, 2008 at 12:11 pm
Polls:
Useful as toilet paper when in hardcopy.
I can’t see anything else they might contribute to society.
November 3rd, 2008 at 2:10 pm
Uh, information?