Split Ticket

Methodology

• All RV/LV polls from nonpartisan pollsters with a FiveThirtyEight rating of B- or better to be included in aggregate
• Partisan pollsters and pollsters with a FiveThirtyEight rating below B- excluded
  • note that bipartisan polls are accepted, such as the ones jointly conducted by Impact Research (D) and Fabrizio, Lee & Associates (R). However, a poll from either of them individually would not be considered.
  • Pollsters with provisional grades of B/C are accepted, as there is at least a B in some part of the rating.
  • In the case of a joint collaboration (e.g. Fabrizio/Impact) with no 538 rating, the poll will be accepted if at least one of the agencies has an individual rating of a B- or better.
• RV Polls receive 3/4 the weight of LV polls
• 30% of the average is the raw average of polls whose last day in the field was within the last 7 days
• 50% of the average is the raw average of polls whose last day in the field was within the last 14 days
• 20% of the average is the raw average of polls whose last day in the field was within the last 21 days
• If an agency polls more than once in a window for the same sponsor, only the most recent poll of the agency/sponsor pair is taken, so as to reduce the influence of trackers and not double-count them.
• Data scraped from FiveThirtyEight

Note: This is not our forecast; those use many more things than a simple generic ballot average. This is also not a model. This is simply a tool that we have built to better inform readers and election watchers in a transparent and open manner about what nonpartisan pollsters say the electorate will be like in the coming cycle.

Data Table

If you would like to see what a high-quality aggregate that considers nonpartisan and partisan polls alike looks like, please visit FiveThirtyEight to look at their average. If you notice any errors, please DM @lxeagle17 and @SplitTicket_ on Twitter.

Ed. Note on the table: The Economist’s G. Elliott Morris recommended that the YouGov/Economist polls of May and June be disregarded due to a programming error in the polls that meant that the results were not actually proper generic ballot polls. Accordingly, as per the recommendation of The Economist’s lead correspondent, we have removed the affected polls due to the bug identified in the programming of the survey.