2016 Election Night

Unemployment rate (September): 5.7 percent (U.S.: 5 percent) Median income (2015): $55,702 (U.S.: $55,775) Poverty rate (2015): 13.2 percent (U.S.: 14.7 percent) College share (2015): 29.7 percent (U.S.: 30.6 percent) Pennsylvania wasn’t hit as hard by the recession as states such as Arizona, Florida and Michigan — its unemployment rate never reached 9 percent in the recession — and it saw a relatively strong recovery. More recently, however, that progress has stalled; the state’s unemployment rate has risen by a full percentage point since the end of last year and is now solidly above the national mark. Really, though, it is a mistake to think about Pennsylvania as one state. Philadelphia and its suburbs are solidly part of the Northeastern corridor; Chester County, west of Philadelphia, is one of the richest counties in the country. But western Pennsylvania has never fully recovered from the decline of the steel industry; even Pittsburgh, which has experienced a revival in recent years, is still losing population.

How do you catch a cloud and pin it down? Back in 2008, researchers in Italy and Canada proposed an idea that makes a hell of a lot of sense, at least on the surface. Maybe undecided voters have already made up their minds, but just unconsciously — they haven’t quite made explicit decisions about things they already believe. By that theory, you might be able to predict their votes by looking at the results of the same kind of speed association tests used to measure implicit racial bias. They tried it out and found that, while spoken beliefs were good at predicting the behavior of decided voters, these unspoken implicit beliefs were better at predicting what undecided voters would do. And, lo, the pollsters rejoiced. Until, that is, somebody tried to replicate the study in the context of an actual election. A different team of scientists tried to use the implicit bias technique to predict how undecided voters would vote in the 2008 U.S. presidential election and the 2009 German parliamentary elections. In both cases, the theory failed. In fact, their research found that implicit beliefs were a better predictor of decided voters’ choices, rather than undecided ones’. (Newer research has since turned up results that don’t match either study.) Why all the disagreement? The authors of the 2012 study think it might be because different kinds of elections have different psychological implications. In a presidential election, for instance, voters have probably thought about the politics involved, asked and answered questions to themselves, and established some kind of personal rhetoric. In other words, their implicit and explicit beliefs might be more likely to match.

A month ago, we took a look at where the two candidates had the most “upside” in this election, relative to 2012. Polls show that Trump will likely outperform Romney among non-college-educated whites, and Clinton will outperform Obama among college-educated whites and all non-whites. We hypothesized a scenario where:

• One in five non-college-educated whites who went for Obama in 2012 would switch their vote to Trump.
• One in five college-educated whites and non-whites who went for Romney in 2012 would switch their vote to Clinton.

Here’s how 2016 would look in this “Vote Swap” scenario:

In our final general election polls-only forecast, this outcome looks like a very real possibility. We give Trump a 70 percent chance in Iowa and a 65 percent chance in Ohio. Hillary is a narrow favorite in North Carolina at 56 percent. Every other state is projected to vote the same way it did in 2012 and Maine’s second district is a nail-biter — we put Clinton at 51 percent.