Election 2016



If you’re new to my fearless forecast of the coming election, here’s what I do: I begin with the popular vote, then use statistical relationships that I’ve derived from past elections to translate the popular vote split into electoral votes and changes in the composition of the House and Senate.*

11/03/16 — I caught Reuters cheating (as discussed below), so I no longer use that poll in computing my baseline forecast. The baseline forecast still portends a victory by Clinton, though her lead is shrinking:

  • Clinton takes 51 percent of the two-party popular vote, as against 49 percent for Trump.
  • Clinton wins 276-312 electoral votes, leaving Trump with 226-262.
  • Given the Trump-Clinton split (which isn’t yet a given), the GOP will lose no more than 4 House seats, retaining a solid majority of at least 243-192, though a loss of as many as 16 seats (for a 231-204 split) isn’t out of the question.
    • And given the same Trump-Clinton split, the GOP might not lose a Senate seat, leaving that chamber with 54 Republicans and 47 Democrats (counting the so-called independents as Democrats). However, a 2-seat loss is strong possibility. That would leave the GOP with 52 seats to retain a nominal majority. But the defection of 2 RINOs would leave the Senate tied at 50-50. And if Killer Kaine becomes vice president, his tie-breaking vote would hand control of the Senate to Democrats.

Now, the big picture. The scale for polling results is on the left axis. Additional indicators are measured on the right axis.**


The key events represented by vertical black lines are the first Trump-Clinton debate on September 26, the release of the infamous “Trump tape” on October 7, the second debate on October 9, the third debate on October 19, and James Comey’s announcement on October 28 that the FBI had re-opened the investigation into Clinton’s e-mails.

Just how far south (for Clinton) will things turn? To get a handle on that question, I’ve plotted some polling results since the third debate:


The points plotted at November 8 represent the linear trend in each poll since its most recent peak. The trend lines fit the actual and projected plot points.

Compare the Reuters values with those that I plotted yesterday. Caught cheating for Clinton, and thereby ejected.***

So the valid trends all point to a win for Trump, albeit a narrow one in the case of the RCP 2-way poll. My gut feeling (as of now) is that Trump’s margin of victory in the two-party popular vote is unlikely to exceed 4 percentage points. And he could still lose. Projections (like regression analysis) are accurate only in their representation of the past.

Where will it end? Stay tuned.
* I start by averaging the current split between Trump and Clinton in these polls and aggregations of polls:

  • the Reuters poll, which is heavily skewed toward Clinton, but which I’ve adjusted to the  account for the likely direction of respondents who now say that they’ll vote for Johnson, Stein, or “other,” or who respond “wouldn’t vote” or “don’t know”
  • the two-way (Clinton vs. Trump “poll of polls” at RealClearPolitics (RCP), which I adjust as discussed in this post
  • RCP’s 4-way poll (Clinton, Trump, Johnson, Stein), similarly adjusted to account for likely defections from voters who say that they prefer Johnson, Stein, or “other”
  • and, for balance, the IBD/TIPP poll, which has a good track record, a high rating from FiveThirtyEight, and is somewhat of an outlier in that it’s less favorable to Clinton than the preceding polls. (I’ve also adjusted this poll to account for the likely direction of respondents who say that they’ll vote for Johnson, Stein, or “other,” or who respond “not sure.”)

** In addition to the Reuters, RCP, and IBD/TIPP polls (see preceding footnote), the graph includes the USC/LA Times poll, which is another Trump-leaning one. All of these polling results are plotted on the left axis.

These are the additional indicators, plotted on the right axis:

  • the Iowa Electronic Markets (IEM) Winner-Take-All (WTA) market, where the IEM WTA line represents the percentage-point spread between the percentage of money bet on Clinton and Trump
  • Rasmussen’s approval index for Obama (percentage of respondents strongly approving of his performance minus the percentage strongly disapproving), which I report because perceptions of Obama’s performance are likely to rub off on Clinton.

I plot all of the values against the dates on which polling was conducted or bets were made, not the dates on which results were released. And in the case of multi-day polling, I use the central date of the polling period. Therefore, almost all of the indicators are slightly out-of-date, a fact that one should consider when interpreting the indicators — especially if the race continues to tighten.

*** I said this yesterday:

I must draw your attention to the downward trajectory of the Reuters poll. Of the polls that I track, it has been and continues to be the most favorable to Clinton.

The following graph, from yesterday’s version of this post, is the one that inspired my statement:


I derived the values for the Reuters poll from results that appeared briefly online and then were withdrawn. New values, much more favorable to Clinton appeared this morning and are included in figure 2 (above).


Comment at your own risk.

Please log in using one of these methods to post your comment:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )


Connecting to %s