A Tool To Help Study the Intersection of Poll Methdology and Election Poll Accuracy

The Experts on Google Consumer Surveys - 02/20/08 03:40 PM

How can you tell the difference between one election poll and another? Here’s a start … (keep clicking on the graphic until the size is legible).

FL GOP Primary 012908 All Pollsters x All Measures

Your help filling-in the handful of¬†empty cells (flagged with¬†“?”), is appreciated.

The goal of this undertaking is to better understand the correlation between the methodological choices an election pollster makes and the results an election pollster produces.

This chart permits an analysis of variables such as: 

  1. Significance, if any, of the length of field period.
  2. Significance, if any, of the proximity of poll release to election.
  3. Significance, if any, of the number of undecided voters.
  4. Significance, if any, of the number of respondents interviewed.
  5. Significance, if any, of the source of the sample.
  6. Significance, if any, of the type of interviewing technique.
  7. Significance, if any, of the method of respondent selection within the household.
  8. Whether one error-measure produces a different “most accurate” pollster than an0ther.
  9. Whether one error-measure favors certain types of pollsters over others.

The results from the Florida Republican Primary, shown here, are not intended to be illustrative of anything other than a high-profile contest in which a lot of pollsters released research during the 7 days immediately prior to the vote. Results from other primaries, to be released over the coming days, may show other pollster rankings.

Suggestions to: editor@surveyusa.com

Copyright 2017 SurveyUSA®, Clifton NJ, all rights reserved. Terms & Conditions.