Maya Benowitz, Mathematical Physicist/Biophysicist
Written 20 Mar
I would equate the accuracy of his 'model' to flipping a slightly biased coin for every election and getting heads every time. If I had $10 million to wager against his prediction, I would do so confidently. The 'statistical model' is political pixie dust, lacking any and all flavors of rigor. There is a fundamental problem in building predictive models from historical events. The further you go into the past, correlations exponentially decouple with respect to the present. This exponential decoupling is proportional to the exponential growth of technology. The political landscape has fundamentally changed in the post-internet era, any model not accounting for this change almost surely will fail.
How statistically sound is Helmut Norpoth's 'primary model' for predicting presidential election results? - Quora