Decoding the polls – The difference the sample makes
Bottom Line: Polls...love them, hate them or both depending upon what they say. It’s important to take all of them with proper context in mind. The first dynamic to consider is the sample of the polling. From time to time when discussing President Trump’s approval ratings previously, I’ve highlighted the difference the sampling makes. In the case of his presidency it’s been significant. That remains the case today. Look at an average of all national polls over the past two weeks and you’ll see this average approval rating for President Trump: 43%. Maybe that’s representative of his approval rating across the country, maybe it isn’t, but here’s where it’s not instructive...the 2020 Presidential Election outlook. Yet isn’t that the way approval ratings are commonly positioned in the news media? This is why sample matters. There are two ways with which samples should be considered if one’s trying to gauge election outcomes. Partisan makeup and level of engagement. In this story I’ll focus on level of engagement. There are three types:
- Adult samples
- Registered voters
- Likely voters
Adult only samples are the least reliable as they include those who aren’t even registered to vote along with those who are. Better is polling among those who are at least registered. Best are those who aren’t just registered but likely to vote. Remember that 43% approval for President Trump? Here’s the real story behind that number. Using the same polls that comprise that percentage here’s the difference based on sample...
Trump approval percentage:
- Adults: 41%
- Registered voters: 42%
- Likely voters: 46%
Notice President Trump’s approval rating rises with each level of engagement. Now why is this especially significant in the context of a Presidential reelection bid? Consider that on the exact date in his successful reelection bid President Obama held a 45% approval rating with likely voters and a 47% overall approval rating. That’s right, President Trump’s average approval rating is 1% higher among likely voters at this stage in the cycle as compared to his predecessor. Now, compare that to the media’s coverage of polling. And that’s the current example of the difference samples make.