According to numerous political pundits (and certainly pollsters), the outcome of the 2016 election was the biggest political upset in modern history. Perhaps... but should Trump defy the pollster odds and defeat Biden on Nov 3, the shock would be even greater. In fact, according to Deutsche Bank's Jim Reid, a Trump victory would be "the biggest polling shock in history."
Let's back up: with Joe Biden seeing his lead in various polls rise by over 10pp over the past week, a fact both the liberal media and various Gen-Z traders have pounced on to predict a landslide victory for Biden and even a Blue Sweep so powerful, it eliminates a "contested election" as a possible outcome...
... Jim Reid decided look at how big the polling errors were in previous post-WWII elections.
This is shown in the graph below which compares the final poll or polling averages (since 2004) for each election versus the eventual vote margin. Polls have been within 0-3% of the final outcome in the last 6 US presidential elections. And though polls have (far) more often erred away from the Republican candidate over this period - indeed, in the past seven elections, polls have "oddly" shown a bias for the Democrats on 6 out of 7 occasions - the median error is less than 1%.
The largest error was in 1948 when President Truman won by 5% in spite of being behind by 4.4% in the final polls. However, Truman’s challenger in New York Governor Dewey saw his lead fall from 17 points in late September to 9 points in mid-October before settling at ‘only’ 5 points just before the election.
The last two big misses were a 6.4 point overstatement of Clinton’s eventual 5.6 point win in 1992 and a 6.4 point understatement of Reagan’s 9.4 point win in 1980. These receive less attention, however, since the error didn’t change the result that the polls were already implying.
What this means is that while a Truman-style error in the polls may give Trump a chance given the electoral college system, Reid concludes that "the reality is that - unless the polls narrow into election day - a Trump victory would be the biggest error in our modern era of mass polling."
Which, in light of the media's now overt bias and pollsters' catastrophic track record in recent years, would be an appropriate continuation to the four craziest years in US politics.