Why the huge spread? ( In Gallop polls) The New York Times Sept 30

 

By Nate Cohn

In the wake of Tuesday’s messy debate, we got a different kind of survey: a so-called instapoll. This is a survey of people who tell a pollster they will watch a debate and agree to take a poll after the debate is over. These surveys are quirky in some ways, but they do a decent enough job of helping us gauge the effect of a debate. This time, the news was generally good for Joe Biden.

Why the huge spread? Instant-reaction poll samples can be pretty biased. Often, you’re looking at respondents who have negotiated three steps: They agree to take a poll before a debate; they agree to take a poll after watching a debate; they actually follow through and participate in the poll after the debate. They’re also people who have watched a debate, who aren’t representative of the overall electorate.

On top of that, pollsters have to decide whether or how to undo those biases. All of this can lead to a huge spread in the results. You might be tempted to just discount these instant poll results altogether.