Nate Silver’s analysis of poll-herding leads me to trust polls less than ever, yet, I have come to believe that it’s a fool’s errand to bet against the average of polls. This year, the correct bet was to assume that virtually all the polls were way too favorable for the Democrats, but there really wasn’t any way to guess that they would be slanted that way rather than in the other direction. In fact, my bet was wrong. Based on the available information, I thought it was very unlikely that the polls were biased against the Democrats.
My predictions were based on the assumption that the average of polls was pretty accurate but was modestly undercounting Democrats. This led me to predict that Democrats who were narrowly leading in the polls would, in fact, go on to win. And, in at least one case, it led me to predict that a Democrat who was losing would nonetheless win.
I had no way of knowing that the pollsters were deliberately making the races look better for the Democrats in order not to have their polls look like outliers.
