Much like the news we consume, political polling is often skewed and is less about taking the pulse of the electorate and more about creating a narrative. Also much like the news, where there is consistent results overlap is typically where the accurate information may be.
Over the years, my personal mantra during election season is all polls should be taken with an entire shaker of salt, but lately I have been shifting more towards brining; it’s a far more efficient use of salt and longer lasting.
2026 being midterm elections year means voters are going to be harassed non-stop by pollsters. Some of you will be blocking those calls, others will answer, and some will even lie throughout a poll — well, just because.
That activity or lack thereof means some polls may be accurate, many will be outright wrong, while others look promising but have their own special tweaks hidden in the details. We’re going to look at some special tweaks today.
On X, I follow an account called Interactive Polls, which shares various polls both national and targeted to North Carolina and is generally a useful account worth a follow.
On Feb. 20, Interactive Polls shared a poll by Change Research, which is a polling outfit that partners with NC Democrats and progressive groups.
Change Research doesn’t just work with Democrats, it boasts working to help them nationally. Check out the company’s 2024 “Accuracy” report, scroll down and read the “highlights“ about helping Democrats win over Republicans and its efforts to push movable voters to Kamala Harris in 2024.
The Poll
The Change Research poll was conducted from Jan. 31-Feb. 4 and surveyed 1,069 registered voters in NC. The poll was commissioned by former statehouse legislator Sam Hunt, a Democrat.
Key results included Roy Cooper leading Michael Whatley 50-40 overall and Democrat Assoc. Justice Anita Earls narrowly led Republican Rep. Sarah Stevens 45 to 42 percent, well within the 3.1% margin of error.
These two results are not unlike other poll results over the past 3-4 months and represent the “overlap” mentioned at the onset of this article. Unlike other queries in the poll, the questions posed for both of these races was done in a neutral manner.
Some of the most interesting results of this poll were the responses given by participants in the “Open Ended Questions” section that can be viewed in the crosstabs.
According to the poll’s summary presentation, below is the demographic landscape of the participants; note the 2-3 point higher shift of independents to Democrat when independents with a lean towards the party are added.
The means by which participants were contacted, the areas sampled, oversampling of one party or another, and the way questions are asked all impact the results.
It’s important to note Change Research’s website says it uses “Magnify AI Targeting” to find “the exact individual voters you need to reach,” and the methodology tells us Change Research targeted and recruited participants using specific social media apps and cell phones.
That recruitment was not balanced geographically. The sample was weighted demographically, but also by “2024 presidential vote, and political region.”
Yet the poll oversampled in the high population regions, which are also the bluest areas of the state which broke for Kamala Harris in 2024: Mecklenburg County /Charlotte area and Wake County/Raleigh - Durham area.
Additionally, after having a respondent identify the party ‘they think of themselves’ as being, there is a later question (below) giving participants an option to “choose more than one.” This could be considered a nudge question, meaning it gives the participant the idea of considering another party trajectory when answering additional questions.
More To The Story
The way questions are asked changes how people answer. The more loaded the question, the answers will typically skew harder to one direction over another.
Now that the participants have been given permission (in Question 10) to identify as multiple political persuasions, the phrasing games can begin.
As an example, the questions asked about Immigration were incredibly biased and started by asking two tone-setting questions of only half of the participants, but it doesn’t say which half:
“The power of the federal government is absolute and should never be questioned.”
“When it comes to immigration enforcement, the power of the federal government is absolute and should never be questioned.”
Participants were then asked (in this order) how effective ICE enforcement was in NC, asked if they agreed with the statement, “North Carolina is overrun with criminal illegal aliens. The federal government must do whatever it takes to remove them,” and whether they believe ICE should be “abolished.”
Then came these two questions, which are without argument loaded and lacking any kind of context:
“Recently, in Minnesota, an ICE officer shot and killed a woman in her car and a man on the street. Do you think these shootings were justified or not?”
Do you think ICE officers who kill U.S. citizens during ICE operations should have immunity from any investigation or prosecution?”
Like most statistics, polling data can be manipulated various to get a desired outcome. Historically, with a few pollster exceptions, we’ve seen polling blow it when it comes to underestimating Donald Trump.
This exercise hopefully was useful for folks in going beyond reading the bullet points the poll’s authors want you to see. When present, always read a survey/poll’s supporting documents, such as this poll’s toplines, crosstabs, and methodology.
Parting thought:
Help More To The Story with commissions earned through an Amazon purchase.
Today’s Deals:
Waterpik Cordless Pearl Rechargeable Portable Water Flosser
Snow Joe Ice Beater Calcium Blend Powerful Ice Melt (50 Pound Bag)