While the current daily glut of political polling results has become a popular news destination, it has also evolved into a major source of frustration amongst the public. Candidates, political parties, and supporters unsurprisingly single out and focus on the polls claiming that their favorite candidate is presently in the lead. There’s a lot of speculation and discussion taking place around who’s really ahead and why, and limited attention or dialogue around the reasons various polling results on the same day can differ as much as they do. The disparities in the results depend a great deal on the differences in the research methodologies used to reach them. So comparing results from different political polls can be like comparing apples with oranges; they’re both fruit, but to reach their distinctly different results in flavor, texture, and appearance requires variations in the biological processes to get there.
We examined five well known polls conducted October 23 – October 24, about the upcoming presidential election. It is striking to see that while three of the five polls have Clinton and Trump virtually tied (tied or within 1-2 percentage points), one poll shows Clinton ahead by 5 percentage points, and another poll shows Clinton ahead by 12 percentage points:
A number of factors will impact the results from any poll, including: who is polled, the source of the survey audience, when the survey is conducted, how the poll is conducted – online, by phone, or combined online/phone, how the questions are asked, which response options are provided for given questions, whether the data is weighted to be representative of the target population, etc. These factors alone, or in combination with each other, can and will obfuscate the ability to uniformly compare results from one poll to another.
Tip: when reviewing results from polls, whether focused on politics, marketing, health, food, travel, or any other area of interest, be sure to closely examine the methodology or design, to better understand what a poll is reporting. While many of us researchers strive to deliver reliable and robust insights that drive decision-making, is there a need for some kind of standardization in methodology and reporting results when it comes to politics or other sensitive important topics such as health and policy?