I did a study of reading comprehension for my masters thesis in 1981. I used a methodology that had just been introduced by Prof. Daniel Kahneman. I asked people to rate how certain they were about the information they’d just read instead of simply answering yes-or-no/true-or-false type questions. If they read an article about a traffic accident in town that said: "... one of the drivers was taken to the hospital but shortly released" …they rated on a scale of 0 to 100 how certain they were someone was injured. Since the answer may be something less that 100% …the rating scale was more representative of the reading material and turned out to be a better predictor of reading ability.
Flash forward: Arie Kapteyn, a student of Daniel Kahneman, used a similar methodology in the USC/Los Angeles Times Daybreak poll to predict the outcome of the 2016 presidential election. He asked respondents to rate how committed they were to their candidate of choice as well as how likely they were to vote (also on a scale of 0 to 100). He included all respondents (not just those likely to vote). He then factored-in commitment-level and likely-to-vote ratings and found that it predicted the outcome better than the national polls including Fox and CNN. It consistently ranked Trump ahead of Clinton by 4 points in the months leading up to election day. Surprised the hell out of everyone, including the author of the poll, and validated the methodology for predicting political races as well.