❝ A new question added to the University of Michigan’s Survey of Consumers could turn out to be more accurate than ordinary opinion polls in predicting the outcome of the U.S. presidential election.
In June and July, respondents to the monthly survey were asked who they expected to become the next president — rather than who they intended to vote for. The results belie the horse-race nature of the campaign that’s being implied by most polls of voter intent.
❝ 58 percent of the households surveyed by the University of Michigan said they thought Hillary Clinton would emerge victorious, relative to just 37 percent for the real estate and reality TV mogul Trump. That presents a very different picture to aggregations of voter intention; as is shown by Nate Silver’s FiveThirtyEight, which has Hillary Clinton’s chances of winning the presidency at 53.3 percent versus Donald Trump at 46.7 percent.
❝ A report published by Ludwig Maximilians University Research Fellow Andreas Graefe in 2014 found that asking voters who they think will win has proved a better crystal ball than asking them which candidate they themselves are likely to support.
“Across the last 100 days prior to each of the seven U.S. presidential elections from 1988 to 2012, expectations provided more accurate forecasts of election winners and the final vote shares,” he wrote, relative to benchmark methods like intention polls, prediction markets, expert judgments, and quantitative models. “Gains in accuracy were particularly large compared to intention polls: on average, expectations reduced the error of intentions by more than half.”
Wonder how many professional pollsters will take note of this finding?