Shorter former Census Bureau director and current Discovery Institute president Bruce Chapman (author of a book with this blog post’s title) – The Futility of Polling:
Numbers are stupid.
It’s better to predict the outcome of elections based on conversations with some guy on the street, rather than through statistically validated models parameterized with stratified random samples.
Actual concluding line:
Relying on polls is like an intelligence agency resting its judgment on data intercepts rather than “Humint”, the human intelligence gathered by old fashioned spies. We don’t need spies in politics, but some old fashioned, on the ground interviews would give the current picture more vividness, more clarity.
Counterpoint by someone who has actually looked at the merits of humint over sigint:
the proprietary kind of information that spies purvey is so much riskier than the products of rational analysis. Rational inferences can be debated openly and widely. Secrets belong to a small assortment of individuals, and inevitably become hostage to private agendas. …
the proper function of spies is to remind those who rely on spies that the kinds of thing found out by spies can’t be trusted.
And neither, of course, can the Discovery Institute.