Presidential Poll Performance 2000 Error Calculator
HOW THE COMPUTATION WAS DONE
- Number of digits: Percentages reported for the polls were mostly two digit whole numbers. Therefore, all election results were rounded to two digits, as were any polls that reported decimals. The difference between a poll and the election were also whole numbers. Averages were computed to one decimal place.
- Error on lead candidates: The margin between the top two candidates in a poll is subtracted from the margin between the same candidates in the election. The resulting difference is divided in half to get the error per candidate. There was no allocation of "undecided" vote in the polls for this computation.
- Error on third party candidates: For this computation it was necessary to allocate any "undecided" vote reported by a poll. The "undecided" was distributed proportionally to all candidates reported in a poll. The third party candidate error is the difference between the adjusted poll result and the election.
- Average error: Average errors for the margin, the lead candidates and the third party candidate for each election year were computed to one decimal.
The computation of survey error is slightly different than the one used by NCPP four years ago. The results are close, but not identical, to the previous computation. The differences are as follows:
- This computation observes conventions about significant digits.
- This computation does not make assumptions about the distribution of the "undecided" vote reported in some polls for the error on the leading candidates. It was, however, necessary to allocate the undecided in order to compute the error on third party candidates.
- The error on the two leading candidates is the same for all years, whether or not there are significant third party candidates.
- The error on the third party candidate is independent of the computation of error on the leading candidates.
For more information about this and other polling issues, contact the NCPP Polling Review Board Members.