Are Empirically-Based Predictions of Presidential Elections Rational?

I find myself more and more perplexed these days. Contrary to Pat Moynihan everyone has their own facts these days.

As of this writing every econometric model of the 2020 presidential election predicts a solid victory by Donald Trump, in some cases by more than a standard deviation. Ray Fair’s model for the 2016 election was right on the money—it predicted that Hillary Clinton would narrowly win the popular vote.

Ray Fair and Mark Zandi, two of the primary authors of these econometric models, are both Democrats; they cannot reasonably be accused of bias. If anything their bias would be in the other direction. These econometric models are based on facts not opinions, perceptions, or biases.

Are econometric models of presidential elections rational? Are they meaningless? Or is this time different?

1 comment… add one
  • Guarneri Link

    I don’t even know what the econometric models are based upon, or care. I Am familiar with EMs, took a class in econometrics and most importantly know how they can be abused.

    In polling and deriving conclusions I’m much more interested in poll question biasing, sampling error and honest response. Good luck with that.

    Internal polls may reasonably sift through that. Most polls for public consumption are crap.

Leave a Comment