Ipsos’ final numbers on the eve of the 2016 presidential election had Hillary Clinton ahead by a few points, which is where she landed in the final reckoning of the popular vote. However, as we all know, Trump won the Electoral College.
Ipsos’ Clifford Young and Julia Clark revisit their article published in June 2016 in Real Clear Politics, where they gave far better odds to Trump than others did at the time. Indeed, before state-level polling data indicated otherwise, Ipsos was predicting a Republican victory (and even a Trump nomination) as early as 2015. Ipsos’ long-term forecasting models were spot-on, suggesting a very uncertain election but one tipped in favor of the Republicans.
What happened? We believe that the error was in failing to realize just how close the election was throughout. The polling was ultimately right at the national level, but overstated Clinton in swing states. The 2016 polling miss was not so much a data error as one of interpretation.