… According to number-crunching from Christopher Wlezien and Robert Erikson, political scientists at the University of Texas and Columbia University, pre-election polls make for poor predictors until the close of summer in an election year, by which point both parties have held their nominating conventions. In their book “The 2012 Campaign and the Timeline of Presidential Elections”, they explain that candidates’ standing in the polls fails to account for even half of the variance in their eventual vote margins until the spring before the election. At 330 days before the contest—roughly December of the year before the election—polls show virtually no correlation to final election outcomes; today, 467 days before the 2020 election, they are even worse. One could make better predictions by flipping a coin than by looking at the polls. CONT.
Democracy in America, The Economist