The Princeton Election Consortium meta-margin just took nearly a 1% jump in a day. On polling from before the debate.


In eight years of watching Sam Wang’s extraordinary work aggregating polling data at Princeton Election Consortium ( this is one of the bigger single-day moves I’ve seen, particularly this near Election Day, when polling should be growing relatively static. The Bayesian win probability, which I prefer methodologically to the random drift calculation (priors and posteriors at ten paces if you disagree), is up to 97%, which is now outside two standard deviations.

Most notably, the meta-margin is based on state-level polling, and this should be based on surveys completed before Sunday night’s debate. Begins to explain Paul Ryan’s releasing GOP Congresspeople from supporting their party’s presidential candidate.

Reminder: only ever consider reputable polling aggregation as your source for even a semi-accurate snapshot of the state of a race, understand that it is a picture of a point in time. Wang’s meta-margin, though, does a creditable job of smoothing out the variability between and within polls, which is why you may see numbers from individual polls that look dramatically different. Think of it as looking at the full curve versus a single, semi-random point on that curve.

If it continues to climb, it will rapidly get to a point that is essentially irreversible. It’s why Wang himself has all of a sudden become more interested in downballot races. As you should be too, whatever your own party affiliation. Probably fair, if not ever 100% certain, to consider this one yesterday’s (well, the day before yesterday, really) news, and do what you can to help your preferred party regain or retain control in Congress.


I've got a dog. Used to have a fish.


You can post comments in this post.

  • So in a word: itsovah?

    John Lyons. 7 years ago Reply

    • In a word, mostly.

      John 7 years ago Reply

Post A Reply