The Command Post
2004 US Presidential Election
« CNN: Clark Quits The Race | Main | Wesley Clark Offically Out »
February 10, 2004
| How accurate are the early exits?

[Cross-posted at BrendanLoy.com]

So how accurate are those early exit-poll numbers that The Corner has been reporting these last three Tuesdays?

Let’s take a look at the results (primaries only, not caucuses), comparing the initial numbers to the “final” CNN results. (I don’t have the energy to go looking for the real “final” numbers from all the state election agencies… CNN’s should be close enough for my purposes.)

New Hampshire:
Exit polls: Kerry 36, Dean 31, Edwards 12, Clark 12, Lieberman 7
Actual votes: Kerry 39, Dean 26, Clark 13, Edwards 9, Lieberman 9

Arizona:
Exit polls: Kerry 46, Clark 24, Dean 13
Actual votes: Kerry 43, Clark 27, Dean 14

Missouri:
Exit polls: Kerry 52, Edwards 23, Dean 10
Actual votes: Kerry 51, Edwards 25, Dean 9

South Carolina:
Exit polls: Edwards 44, Kerry 30, Sharpton 10
Actual votes: Edwards 45, Kerry 30, Sharpton 10

Oklahoma:
Exit polls: Edwards 31, Kerry 29, Clark 28
Actual votes: Clark 30, Edwards 30, Kerry 27

Delaware:
Exit polls: Kerry 47, Dean 14, Lieberman 11, Edwards 11
Actual votes: Kerry 50, Lieberman 11, Edwards 11, Dean 10

Tennessee:
Exit polls: Kerry 46, Edwards 28, Clark 15, Dean 7
Actual votes (96% in): Kerry 41, Edwards 27, Clark 23, Dean 4

Virginia:
Exit polls: Kerry 48, Edwards 25, Clark 11, Dean 8
Actual votes (99% in): Kerry 52, Edwards 27, Clark 9, Dean 7

Overall, the early exits have provided a pretty darn accurate picture of what the actual results will look like. The only states with major discrepancies have been New Hampshire (the exits showed Kerry and Dean much closer than they were, 5 points apart instead of 13) and Tennessee (where Clark is doing 8 points better than the polls showed).

Otherwise, the candidate’s actual total has been within 3 percentage points of the predicted total in every case, except Dean in Delaware (4 points off) and Kerry in Virginia (4 points off, tentatively). In most cases, the exit polls have been even more accurate than that.

Bottom line, from this limited sample, it seems that if the early exit-poll data on predicted margins of victory should be looked as having a 6- or 7-point margin for error… though always with some caveats.

Interesting.



Posted by Brendan at February 10, 2004 11:36 PM | TrackBack
Comments
Post a comment

Thanks for signing in, . Now you can comment. (Click here should you choose to sign out.)

As you post your comment, please mind our simple comment policy: we welcome all perspectives, but require that comments be both civil and respectful. We also ask that you avoid the extensive use of profanity, racist terms (neither of which we consider civil or respectful), and other boorish language.

We reserve the right to delete any comment, and to prohibit you from commenting on this site, if we feel you have broached this policy. As a courtesy, we will first send you an email noting a violation so you understand the boundaries. This will occur only once, however, and should we ban you from our comment forums we expect that ban to be permanent.

We also will frown upon those who suggest that we ban other individuals for voicing unpopular opinions, should those opinions be voiced in a civil and respectful manner. The point of our comment threads is to provide a forum for spirited though civil and respectful discourse … it is not to provide a forum in which everyone will agree with your point of view.

If you can live by these rules, welcome aboard. If not, then we’re sorry it didn’t work out, and thanks for visiting The Command Post.


Remember me?

(You may use HTML tags for style)