(Commentary by Kyle Kondik, political analyst at UVA’s Center for Politics) One might have done better in predicting the 2016 presidential election, or at least in anticipating the very close eventual outcome, by basing a projection of the national popular vote on the findings of several political science models released prior to the election. These models mostly made their predictions several months in advance of the election and were based on the incumbent’s approval rating, the economy and other “fundamental” factors.