Nate Silver has been quite successful at predicting the outcomes of U.S. elections in the past, something which is described in his book The Signal and the Noise. The book contains some descriptions of the model used, and a blog post of his describes the model used for the 2014 midterm election. These descriptions are more aimed towards the more general public, and are not very specific in terms of what the statistical model actually is (more than just conceptually).

My question is: does anyone know of any more statistically oriented descriptions of the types of model he has used? Be it scientific papers, blog posts, presentations, replications, etc. His book is very interesting, and I would like to learn what type of model and estimation methods we are talking about here.

Edit: Since someone has voted to close, let me clarify what I am asking. Can someone provide a reference to a description of the prediction model(s) used by Nate Silver, which contains a statistical rather than conceptual description? This does not have to be a published paper by Silver himself, but blog posts (or papers) by others are fine as well.

  • 2
    $\begingroup$ Do you really want to keep this question open, given... $\endgroup$ Commented Nov 9, 2016 at 10:09
  • 1
    $\begingroup$ @AntoniParellada Indeed, perhaps even more so! $\endgroup$
    – hejseb
    Commented Nov 9, 2016 at 17:59

2 Answers 2


There's a description of 538's model here. I'm not sure if it's statistical enough for your liking.


Silver has written about his affinity for Bayesian methods in The Signal and the Noise and elsewhere. I found this book, which is teaches a bit about Bayesian inference through Python examples was a good read on a complicated topic:


It's available on the web and as an iPython notebook, but you can also buy a hard/e-book copy and support the author's awesome work.


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.