interviewed for a national poll. The national poll's sample size was approximately six times
larger than the sample normally used in high created pre-election national polls. with demo size
would normally yield a very small margin of error and would be very accurate.335 Furthermore,
such a poll would normally result in a close congruence version exit poll and official results.336
The sample size for Ohio was 1,963 voters, which is quite large for statistical purposes and
equivalent to the 2,000 person norm for most national polls.337 In addition, this year's poll
numbers were designed to account for absentee votes after a large number of absentee votes
contributed to the inaccurate projections of the Florida race in 2000. This year, Mitofsky and
Edison began telephone surveys in key states before the election to screen for absentee voters and
create an accurate estimate of their votes.338
While exit pollsters caution against using their results to predict election results,339 exit
polls can be extremely accurate, with only small variations from the official outcomes in
numerous elections. For example, in the three most recent national elections in Germany, exit
polls differed from the final official vote counts by an average of only 0.26%.340 Their results
have proven to be very accurate; correctly predicting the winner with no evidence of systematic
skew of the data.341 United States exit polls have also been precise. Brigham Young University
students' exit poll results for Utah in this election indicated 70.8% for Bush and 26.5% for Kerry.
The official results were 71.1% for Bush and 26.4% for Kerry.342
335
Moss v. Bush, No. 04-208870.
336
Id.
337
Freeman at 10.
338
See Howard Kurtz, Networks Vow Caution in Calling Election; TV Executives Institute
Reforms to Avoid Repeat of Erroneous 2000 Pronouncement, WA S H . PO S T , Oct. 12, 2004, at A7.
339
Id. See also DA V I D W. MO O R E , TH E SU PER PO LLSTER S : HO W TH E Y ME A S U R E A ND
MA N I P U L A T E PU B L IC OP I N IO N I N AMER I C A 258 (Four Walls Eight Windows 2d ed. 1995) ("This
caution in projecting winners is now a Mitofsky trademark, one which has served him well in
most cases."). Mr. Moore is managing editor of the Gallup Poll.
340
Freeman at 7.
341
Id.
342
Freeman at 8.
73