There is substantial evidence it remains easier to construct a high-quality live-interview telephone poll than an online poll of equal quality. Just as important, it’s easier to judge the quality of a traditional poll than a nontraditional poll.All of this makes it more crucial than ever for analysts to rely on an average of public polls, rather than citing individual surveys. A decade ago, it was fairly easy to tell the difference between a good poll (high-quality survey research required a high-quality survey design: and a bad poll: pollsters rely on statistical adjustments, known as weighting, to correct their initial samples. A high-quality design without proper adjustment might be fatally flawed, as with Gallup’s results during the 2012 presidential election season, or with many state polls in the 2016 election."
At www.nytimes.com
