Every election cycle, poll aggregators get treated as though they’re forecasting the future with precision they never claimed to have. The problem isn’t the polls. It’s how confidently everyone else reports the number after the decimal point.
A well-conducted poll with a standard margin of error is a genuinely useful tool. It tells you, with reasonable confidence, that a race sits somewhere in a range. What it does not tell you, no matter how precisely an aggregator displays the decimal point, is the exact state of public opinion at a specific moment, and it certainly doesn’t predict turnout, which is often the single biggest variable separating a polling average from an actual result.
I think the honest, if less satisfying, framing that coverage should adopt is this: polling averages describe a plausible range of outcomes, not a single predicted number. When a race is reported as ahead by a couple of points, that’s frequently well within a margin where the actual outcome is a toss-up, a distinction that gets lost the moment a headline turns it into a single, confident figure.
The incentive to overstate precision isn’t going away
None of this is really the fault of the pollsters, most of whom publish careful methodology notes that get stripped out by the time a number reaches a headline. The incentive to present a single, dramatic figure rather than a probabilistic range runs through the entire media ecosystem, not just any one outlet.
My own preference, for whatever it’s worth, would be for every polling story to lead with the range and the uncertainty rather than the topline number. It’s a less punchy way to write a headline. It would also, I think, leave readers considerably less shocked and considerably better informed on election night than they currently tend to be.
