A poll average is a simple concept: just average the most recent polls together. However, the reality is more complex. The New York Times’s polling average is a weighted average, giving more weight to certain polls and making statistical adjustments to address biases in different surveys. Despite these adjustments, the poll average is not a forecast but a measurement of current public opinion.
Here’s how it all works:
Why a weighted and adjusted poll average?
Creating a poll average faces challenges such as determining which polls to include, how much weight to give to each poll, and how to handle biases among different pollsters. One approach is a simple average of recent polls, while another involves weighting and adjusting polls based on quality and biases.
While a simple average offers clarity, it can be influenced by low-quality or biased polls. The Times’s polling average takes a more complicated approach by weighting polls based on various factors and adjusting for biases. However, it does not consider external factors like state demographics or past voting patterns.
The Times uses polls collected by FiveThirtyEight and other sources to create its average.
When dealing with conflicting poll results, likely voter results are preferred over registered voters or all adults. The approach to handling independent candidate Robert F. Kennedy Jr. is also explained, with the average focusing on head-to-head results at this stage of the race.
Weighting polls
The Times’s polling average assigns weights to polls based on factors like recency, sample size, number of polls by a pollster, and the pollster’s reputation.
Weighting pollsters
Pollster weights are determined by professionalism, methodology, partisanship, and experience. Pollsters meeting certain criteria are labeled as “select pollsters” and receive more weight in the average.
Adjusting polls
The poll average adjusts for factors like likely voters, shifts in the race, partisanship, and pollster tendencies to provide a more accurate representation of the state of the race among likely voters. These adjustments help account for differences in polling methods and demographics.
