Tuesday, November 15, 2016

Why the polls were wrong: The undecideds

I have been attending Taleb's lectures on "silent risk" this semester, since we now both work for Tandon. Tonight he was talking about how foolish were Nate Silver's efforts to pin a precise number on the election odds, when there was so much volatility. Taleb recommended modeling an election as a binary option, that pays one if your bet comes in, and zero if not. And with volatility so high, the right price for such an option is about .50... so Silver should have been calling things a toss-up all along.

The volatility was created by the vast undecided or "barely decided" population that kept tipping back and forth.

1 comment:

  1. My rule of thumb is to assume that voters who are undecided in the last week or two prior to the election will break hard against the incumbent (or in an election with no incumbent running, the outgoing incumbent's party). If those voters were happy with the existing situation, they would have decided early to stick with it.

    ReplyDelete

Zeno for the computer age

If you wish to better understand Zeno's worry about the continuum, you could do worse than to consider loops in software. Case 1: You...