The opinion polling business has assumed a vast power and importance in American politics, arguably far beyond the influence of the media, activist groups, or really anybody else. Polls, in the aggregate, are fairly effective at what they do, but they also tend to project a false sense of precision in an imprecise business.
Nate Cohn of the New York Times’ Upshot blog has done an enormous public service by a simple experiment: he provided the raw data from a public poll to four different reputable pollsters and poll experts, and they came back with four different answers – reflecting the extent to which even the best of polls incorporate judgment calls, assumptions and preconceptions in addition to simple data.
It’s worth a read, even if you don’t normally enjoy this kind of thing, just to see how the polling sausage gets made. These are diligent pollsters, doing their best and knowing their work will be transparent to Times readers. But their conclusions are still different in potentially important ways.