The latest polls showing Clinton would win the presidential election is more evidence to support my argument that trying to predict human behavior — whether it involves economic decision making or politics decision making — involves dealing with a nonergodic stochastic process.
Pollsters believe that if you take a RANDOM SAMPLE OF THE VOTING POPULATION ON DAY X BEFORE THE ELECTION and calculate the probability distribution of voting for the various candidates, then this x day before election probability distribution is equivalent to the probability distribution you will get from any sample drawn from the same population universe on election day. This presumption would be correct if the stochastic process generating voters preferences is ergodic.
But as I have argued regarding economic decisions — drawing a sample on day x provides no reliable probability distribution of what economic decision makers will do on day x + y.