Last night, the PBS News Hour's Paul Solman tried to highlight the differences between election polling, econometric modeling (in this case, Ray Fair's) and betting markets ("Political Polls, Professors and Election Markets Predict the Presidential Race"). But it took commentator David Brooks, appearing a few minutes later, to explain. These approaches offer competing readings on today's electorate, but predicting the election results is entirely another matter. Brooks did not mention Black Swans, but they can easily materialize.
One would think that of the three, only the betting markets were forward looking. The models are based on historical data, the polls are random chat (who decides to answer the phone and stay on the line), but people putting up cash might be looking ahead (as best they can) as well as backward.
This is another area where the News Hour report was muddled. Polling and betting go on until almost the last minute. They offer many readings along the way. Which of these are scored when evaluated against the actual results? The very last ones.
It's an old story. Forecasting is a fool's errand. Some people must plan and take a stand in some situations. For others, it's simply a parlor game.