Reasons for polling failures in US election are unclear

Nature : Before the US presidential election, the majority of major national polls suggested that Hillary Clinton would win the presidency by a 3–4% margin in the popular vote. Poll aggregators, which apply algorithms to national and state polling numbers to attempt to paint a clearer picture of the outcome, predicted a 71% or better chance of a Clinton victory. Obviously, that did not happen. Clinton ended up, as of 10 November, with a popular vote lead of only around 0.2%, and she lost to Donald Trump in several states she was expected to win. Claudia Deane of Pew Research Center in Washington, DC, says the surprising thing is that so many polls with such a wide range of methodologies all erred in the same direction. Some potential factors contributing to the widespread polling failure include poor assessments of likely voters, people misrepresenting their voting intentions, and poor coverage of certain segments of the voting population. Some of those problems are likely rooted in the shift away from landlines to cell phones, as many people do not respond to unknown phone numbers. Meanwhile, online pollsters often have difficulty obtaining a statistically useful distribution of respondents.