Among the 5 most popular polling firms, including Quinnipiac, SurveyUSA, PPP, Rasmussen, and CNN, the firm with the smallest disparity between its polls of Senate races and the actual outcome of each race is Quinnipiac. On average, each Quinnipiac Senate poll was was within 2.88% of the outcome. Though Quinnipiac had the smallest average error, Quinnipiac only released 16 different polls of 6 separate Senate races – the least of any of these polling firms, as the chart below shows. This means a polling firm with a low volume of polls in comparison to other firms may result with a disproportional and unrepresentative statistic of their polling accuracy. Accordingly, in addition to the average error of each poll, we must consider another factor of polling firms: the ratio of a firm’s poll(s) within 1% of the outcome and the total number of polls of a firm. This will be referred to as the “Close-Knit Factor,” which allows for a more complete comparison between the accuracy of polling firms and the amount of polls each firm releases.
Quinnipiac released 16 total polls and 6 of these polls were within 1% of the outcome, so Quinnipiac’s close-knit factor is 37.5%. After comparing the close-knit factor among these 5 polling firms, Quinnipiac is still the most accurate firm, but the order of the other firms does shift. SurveyUSA drops to 3rd, whereas PPP rises to 2nd. Even though PPP had a greater average error than SurveyUSA, PPP released a greater percentage (12.7%) of Senate polls that were within 1% of the outcome than SurveyUSA. As a result, PPP Senate polls were more accurate than SurveyUSA Senate polls.
The chart below lists the polling firms, the amount of polls each firm released between October 3rd and Novemeber 1st, the amount of separate Senate races each firm surveyed, the average error of each firm’s polls, and the close-knit factor of each firm.