Labor Day is the traditional start of the fall campaign season. Hard to believe since the Presidential race has been on for two years, but the election is less than sixty days away.
This is where polls become the life’s blood of a candidate’s campaign. Based upon the most recent polls, how are Donald Trump and Hillary Clinton doing?
According to the pundits the race has pushed back into the tied region. A conclusion they make based on national polls – but the White House is not decided by a national popular vote but instead by the electoral college. This makes this conclusion interesting in terms of the two candidates’ popularity, but invalid as a harbinger of winning the election.
The second claim of a tight race is based on a misunderstanding of the margin of error. There is a misconception that says when the two candidates are polling within the margin of error of each they are tied – they are not.
With these two disclaimers, how close is the race?
An alternative to drawing conclusions from the individual state polls is to run an election simulation. The model – which uses state by state polls not a national poll – statistically simulates running the November election in each state thousands or millions of times to determine the probability that a particular candidate wins the necessary 270 electoral votes.
Using the most current polls, this model estimates that Secretary Clinton wins at least 270 electoral votes and thus the Presidency 95.7% of the time.
The magnitude of the win is often used an indicator of a mandate. The model predicts that Clinton wins an average of 322 electoral votes, fifty-two more than the needed 270.
The expected number of electoral votes may not be the most common number of electoral votes that she wins. The most common result, the mode, is 332 electoral votes, sixty-two more than 270. This occurs just a bit over 4% of the time.
Both of these values are interesting, but they are but single results – and both with low probabilities of that exact result. Instead, a more useful measure is the confidence interval. This is the centered interval which covers the middle 95% of the model’s outcomes.
At this point in the campaign, Secretary Clinton’s 95% confidence interval is 268 to 390 electoral votes. This does show a very small window in which Donald Trump wins.
The model also brings to light some interesting state wide results.
The must win state for Trump, Ohio, is virtually tied with Trump and Clinton each winning 50% of the time.
Pennsylvania, the other northern must win swing state, goes to Clinton 88% of the time. Not a guaranteed win, but not close enough to use as your base for a Republican electoral college win.
Florida is much closer than PA, but Clinton still shows a command of the probabilities with a 66% chance of winning the state’s twenty-nine electoral votes.
Similarly close races are in Nevada and Virginia with Clinton winning 61% and 62% of the time respectively.
The curious results are the reliably red Utah and Georgia. In 2012, Romney won Utah by forty-eight points, and Georgia by seven points. Today Trump still wins Utah but only 82% of the time. And Georgia is tied.
Volumes have already been written about how this year’s presidential election is unlike any that has come before. These same articles often present the election as a horse race with both candidates within striking distance of edging out the other. And while this election may be many things, a tie is not one of them.