Forecasting elections with October surprises

Abstract

October Surprises are the bane of election forecasting, especially for models that post their forecast early. In 2020, several surprises occurred well before October, beginning with the onset of the Covid Pandemic, followed by the economic depression triggered by the lockdown, and sweeping changes of voting procedures along with racial strife in the wake of the killing of George Floyd. These surprises, the evidence shows, derailed Trump's re-election prospects. In so doing, they also upended the forecast of the Primary Model, posted March 2, 2020, that gave Donald Trump a 91% chance of winning the 2020 presidential election. It was a once-in-a-century miss of the Primary Model. The lesson for forecasting with that model: Update the early forecast all the way until the last primary is over! It is a tribute to Harold Clarke that the main surprise, the Covid Pandemic, is the kind of issue whose electoral power he has tirelessly demonstrated in his studies of voting around the world: a “valence issue.”