How is the RCP Trump Approval polling average constructed, and what factors should be considered in estimating it for the week ended 8/11 (or any week)?
Hi! Thanks for your question about how the RCP average is constructed and the factors that could influence that average one way or another. Here's the short answer: RCP's average is a simple unweighted average of ten recent presidential approval polls, a mix of monthly, weekly, and rolling daily surveys. Several factors can influence the RCP average and could be used to make a prediction on the average at the end of a given week. Read on for my deep dive!
I started my search at RCP's site to see if I could find any information about their methodology. While I found the polls they use and their update frequency, unfortunately, I wasn't able to find any information about how or why they choose the polls they do. I found an article on general poll aggregation methodology that might also apply to RCP, but nothing on or from RCP specifically.
I also looked at sites for each pollster currently included in the RCP average to see their frequency of updates and whether or not they offer subscriptions to get early data. I found data on most of these outlets, but for those who didn't make that information clear I estimated based on their refresh dates in RCP's average timeline.
Lastly, I looked at FiveThirtyEight's approval tracker to see their methodology and compare historical data to the RCP average. FiveThirtyEight was very clear about their methodology, which differs from RCP in several significant ways.
RCP AVERAGE METHODOLOGY AND PREDICTIVE FACTORS
RCP's approval polling average is a "simple unweighted average" of what looks like ten recent approval polls from various sources. The RCP average updates daily, though I wasn't able to discern what time the update goes live. HuffPo's similar average handles daily rolling average polls by updating their total when "all of the data are new" -- every three days for Rasmussen and Gallup, for example. Based on RCP's timeline of past polls, it seems like they may do this as well.
The biggest factor to consider in predicting the RCP average is the news for the week in which the average is calculated as it coincides with poll timing. Several of the polls have regular survey intervals; studying the news during the times when surveys are happening could help indicate which way the average's needle will move.
The presence of Rasmussen data in the polling average is another key factor in predicting the RCP average. Rasmussen has been frequently cited as a high outlier in the polling averages since the early days of Trump's presidency -- their Trump approval numbers have been consistently higher than other shops. According to HuffPo, "no other pollster" has as much influence on polling average variance as Rasmussen. If the RCP methodology follows the HuffPo model of recalculating when the data from a rolling average poll is all new and if the Rasmussen three-day average is due to reset at or near the end of a week, the poll could have a significant effect on the RCP average. Rasmussen does seem to be one of the few polling outlets that offers a subscription for earlier poll results, though I wasn't able to confirm that for certain.
Click here to see my spreadsheet of info on the polls that RCP seems to include. Where possible, I tried to pull frequency and release-time data from the polling outlets themselves, but for several outlets (CNN, CBS News, Fox News, and PPP) I had to estimate based on the frequency with which they appear on the RCP average timeline.
RCP AND FIVETHIRTYEIGHT
RCP and FiveThirtyEight's polling averages both update daily. However, the FiveThirtyEight average is weighted according to their pollster ratings and includes many more polls than RCP's average. Due to these and a few other factors outlined on FiveThirtyEight's site, their average seems to sit a hair lower than RCP's historically.
RCP's average of Donald Trump's presidential approval ratings updates daily and gives an unweighted average of approval figures from 10 polling outlets. The presence of new Rasmussen data in the average, in addition to the news of the week, could be key in predicting the RCP average at the end of a given week. This spreadsheet contains information on the polls included and their frequency -- only Rasmussen and Gallup seem to offer subscription services that would allow early access to results.
Thanks for using Wonder! Please let us know if there's anything else we can do to help.