I like market-based predictions because they can take all of the information that everyone already models and add three things. First, breaking information that is not yet a data point for a model. Second, known information that has no history, thus hard to model. Third, dispersed information that is not widely known.

First, PredictWise went up 5 percentage points during the first debate and subsequent 5 percentage points as the week unfolded. Reliable poll-based forecasts, like the New York Times’ Upshot, went up 6 percentage points, but did not even start moving for three days. FiveThirtyEight went up 15.5 from its midday low of 51.5 on September 26. But, that swing did on start in earnest until three days later as well.

Second, there is no way to model the impact of Clinton having a much larger Get-Out-the-Vote operation in the key swing states. It has never happened before, so there is not sound historical data for this. But, market-based forecasts have Clinton systematically ahead of polling-based forecasts in key swing states. This is likely due to some combination of GOTV, expectations over the two additional debates, and any other fall-out from Trump between now and Election Day. Any difference between poll-based and market-based forecasts on Election Day would be a good quantification of GOTV alone.


Third, anyone is free to include inside information in the markets. I have no basis for it happening this cycle, but wait to see if there is any run-up on pricing prior to the next big October Surprise!