Did Secretary of Fracking Hillary Clinton lose Michigan in part because she supports the economic model that poisons water at home and abroad? Maybe. The connection struck me (as I learned later it had others).
But Bernie Sanders’s primary upset in Michigan, busting many a prediction, including FiveThirtyEight’s pooled polling model, clearly unveiled loaded epistemologies.
A number of oppositional wags–funny but accurate–have characterized polling in the U.S. as testing whether voters have adequately assimilated ruling class propaganda. That is, at heart all polls are push polls, part and parcel of managing expectation. Until they don’t.
A couple years ago, suffering a case of World Cup fever, I wrote three short paragraphs on Nate Silver, pointing out the Bayesian on which his shop depends can be saddled with the premises of power, a problem far beyond mere technical disagreements.
The politics of the nomenklatura–rolling over systemic prime directives one time interval to another in the face of what it dismisses as stochastic noise–can be found down in the very mechanics of statistical modeling.
But what happens when the rest of the world begins to make itself matter? Do the models break down from more than just bad priors and missed predictions?