Last week’s CNN poll showed a tightening race. Why? The poll did not over-sample Democrats as bad as previous polls. The poll did not under-sample Republicans as bad as previous polls. In short, the sample size almost reflected the turnout ratios of Republican, Democrats and independent voters. So why the sudden improvement in accuracy?
Could the current poll be manipulated to give the appearance of a larger post-convention poll bounce? Specifically, when they release the first CNN poll after the convention, will they broadcast the poll in relative terms to fake a larger than life convention poll bounce?
They will use the relative poll numbers to parade headlines such as:
- Huge Convention Poll Bounce for Biden – CNN poll shows big spike in support for Biden
- Poll Shows Voters Flock to Biden After Convention
- Etc.
I’ve been following the CNN and other polls for months now, here is how they manipulate.
- CNN manipulates their poll numbers by controlling the ratio of Democrat, Republican and independent voters in their sample.
- They either lower the number of Republican or Independent participants, or increase the number of Democrat participants.
- Controlling these ratios produces very predicable outcomes.
Last week’s tightening CNN poll had this as a sample pool:
Among the entire sample, 31% described themselves as Democrats, 27% described themselves as Republicans, and 42% described themselves as independents or members of another party.
Three things:
- This ‘tight poll’ had the fewest percentage of registered Democrats after looking at the last 5 polls.
- This ‘tight poll’ had the largest percentage of registered Republicans after looking at the last 5 polls.
- This ‘tight poll’ was the closest CNN has been to actual 2016 turnout ratios by party ID. So, ironically, this was the most accurate CNN poll because the sample size of Republicans, Democrats and independent voters was the closest to actual 2016 election ratios. (Note: CNN also does a lot of demographic oversampling as well, but for this analysis I’m only focusing on party ID).
In the 2016 presidential election, turnout by ID looked like this:
Turnout by Party ID 2016 | clinton | trump | other/no answer |
---|---|---|---|
democrats 36% | 89% | 8% | 3% |
republicans 33% | 8% | 88% | 4% |
independents 31% | 42% | 46% | 12% |
Previous CNN Polls: The ‘tight poll’ was the closest to actual turnout numbers for 2016. It’s clear that when CNN’s sample is closer to reality, their poll ‘tightens.’
- 8/16 poll – poll’s sample was made up of 27% Republicans. 31% Democrats. It showed Biden +4.
- 6/8 poll – poll’s sample was made up of 25% Republicans. 32% Democrats. It showed Biden +11.
- 5/13 poll – poll’s sample was made up of 26% Republicans. 34% Democrats. It showed Biden +10.
- 5/2 poll – poll’s sample was made up of 26% Republicans. 33% Democrats. (Before Biden nomination)
- 4/9 poll – poll’s sample was made up of 25% Republicans. 35% Democrats. (Before Biden nomination)
Prediction:
CNN’s next poll will swing back to their old ratio of Democrats, Republicans and independent voters. Because of the over-sampling and under-sampling, Biden’s numbers will go back to +10 and they will declare the convention a success.