The bigger problem for polling is the specific nature of the error, as that destroys the usefulness of polls. If you have samples that are representative-ish, you can have reasonable error margins around your polls and it's approximate, but so be it. If you literally cannot get people to respond, that is a different kind of error. It's highly directional, and it's correlated across all the polls. Worst of all, you don't know how big the error is (you can guess, of course, usually by fitting to past results), but you don't actually know until after the fact and you have no good way of estimating it. So to that end, I think the intelligent critique is not that the industry is in a coma, but rather that it faces a truly intractable problem of subgroups of people simply refusing to be polled, and we don't have a great estimate of the difference between the polled and unpolled, only that it's not stable over time. If you adjust, are you now too Democratic? Too Republican? Too... neither and just weird? This sort of uncertainty matters most in close races, of course. In a very wide race, fine, D+20 is not becoming R+1. But D+4? R+2? You might as well throw darts at that point. The core issue is not that pollsters are not aware of it. It's that, for this specific kind of statistical uncertainty, it's far harder to do anything useful with it and they are somewhat to completely powerless.