In his first 2016 simulation, he assumed that each group divided their votes (according to state-level exit polls) identically as they did in 2012. Not surprisingly, (since they won comfortably in the Electoral College in 2012), Democrats would also take the presidency in 2016 under this scenario, adding North Carolina to the states Obama won in 2012.

In the second simulation, Oakford assumed 2004 performance levels by both parties (giving the 2016 Republican nominee the benefit of George W. Bush’s relatively strong performance with Hispanics). The race gets closer, but the Democrats win. 

In the final simulation, Oakford gave Republicans the benefit of their high vote share among whites in 2012 and their relatively high vote share among Hispanics in 2004. Even with these favorable assumptions, the Democrats still get over 270.

Oakford’s key assumption is that turnout rates by ethnic and racial groups will stay the same. That’s a big assumption. As discussed above, net votes generated is the metric that matters. Because of their near-uniform support for Democrats, increased turnout by black voters translate on an almost one-to-one basis with increase in net Democratic support. The number of eligible Hispanic voters is certainly increasing, but their turnout rates are low, and, while they strongly support Democrats, they do not uniformly do so. Furthermore, Hispanic voters are more likely to live in uncompetitive states like California, Texas, and New York.