US Presidential Election – Picking winners – Nate Silver and 538

Write these numbers down: 314.6, 50.9% and 52.5.

Nate Silver is a statistician, sabermetrician (baseball analyst), psephologist (election analyst) and writer.  Just like Antony Green, without the baseball.

Nate’s blog is “FiveThirtyEight: Nate Silver’s Political Calculus” and contains the essence of his statistical methods with words to help us along.

I started reading 538 (the name was taken from the number of electoral college votes) before the 2008 presidential elections and Nate was scarily accurate, correctly picking the winners of 49 or the 50 states; only spoilt by Obama’s win 1% in Indiana.

In Australia, the popular vote is sometimes given too much attention since it is the number of seats that determine the parliament.  Similarly, or perhaps more so, the popular vote in the USA has little to do with the result, it is the person who gains the most electoral college votes.  All that you need is a majority of votes from people in the state and you’ll get all of the electoral college votes of that state in most cases.  (Actually, with first-past-the-post you only need more votes than the other guys, not a majority.)

Nate ranks polls by accuracy and bias (by design, by accident or on purpose) assigns weightings and adds his own findings to present  analysis with commentary.  A poll can be biased because it only contacts landlines or only asks “likely voters” (which in Australia would be everybody), both of which may exclude young adults and busy people thereby not polling a cross-section of views.

Also, the sheer number of polls means that Nate Silver can comment put the pundits pronouncements to the test.

538 is predicting a 91.6% chance that Obama will win; it was as low as 61.1% on 12 October.  If Obama is predicted (by 538) to get 50.9% of the popular vote, why does he have such a high probability of winning?  By running a lot of simulations, Nate Silver can see how many times a particular set of results occurs, such as:

  • 0.2% Electoral College tie (269 electoral votes for each candidate)
  • 5.7% Recount (one or more decisive states within 0.5 percentage points)
  • 87.3% Obama wins popular vote
  • 12.7% Romney wins popular vote
  • 0.6% Obama wins popular vote but loses electoral college
  • 4.9% Romney wins popular vote but loses electoral college
  • 0.3% Obama landslide (double-digit popular vote margin)
  • <0.1% Romney landslide (double-digit popular vote margin)
  • 0.1% Map exactly the same as in 2008
  • <0.1% Map exactly the same as in 2004
  • 99.7% Obama loses at least one state he carried in 2008
  • 4.3% Obama wins at least one state he failed to carry in 2008

As at 04:00 ET 6 November (23:00 AEST 6 November) Nate Silver predicts that Obama will win 314.6 electoral college votes, 50.9% of the popular vote and will have 52.5 senators.  We’ll start to see in about 12 hours.

I wonder if he’s met Antony Green?


Your thoughts?

Fill in your details below or click an icon to log in: Logo

You are commenting using your 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 )

Google+ photo

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

Connecting to %s