The Los Angeles Times had a poll which was interesting because I was always up in that poll. They had something that is, I guess, a modern-day technique in polling, it was called “enthusiasm”. They added an enthusiasm factor and my people had great enthusiasm, and [Clinton’s] people didn’t have enthusiasm.Donald Trump (Source: NY Times)
One of the many bizarre aspects of the 2016 US Presidential Election was that so few people correctly predicted the outcome from polling data, including statisticians at media outlets like FiveThirtyEight and the New York Times. Although these outlets consistently predicted a Clinton victory, they disagreed significantly about the likelihood of that victory (see figure below).
STOP: Note in the figure above that the two charts appear to have the same shape over time. Why do you think this is, and what do you think caused the projections to fluctuate?
When we observe differences between election forecasts, it should give us pause about the reliability of these forecasts. It also begs the question, “What caused the discrepancies?” To find an answer, we wonder how we might forecast an election from polls in the first place.