No doubt political analysits will assess why the polls were so far off. But beyond polling errors there were several other reasons why this came as a much bigger surprise than it should have:
Probability estimates are a horrible invention
One of the worst inventions of the 2016 campaign was the “probability model” in polling – the idea of showing the odds or likelihood of a certain outcome. The New York Times got much attention for estimating that Hillary had an 85% chance of winning.
This approach dramatically magnifies the significance of small edges in support. If you are consistently up by 3% then you have an 85% chance of winning – even though a shift of a few points in the other direction would drop the probability to a number like 30 percent.
When it came to race, we mistook trend lines for raw numbers
We’ve heard so much about the browning of America that we forgot some basic math. A massive increase in the number of minority voters would be swamped by a slight increase in the number of whites. Whites may be declining as a percentage of the electorate but they’re still roughly 7 times more numerous than Latinos.
People on the left have long complained that those on the right were in an information bubble. Turns out progressives were in one as well. They had a completely warped sense of their own strength because Facebook feeds us the most “relevant” news, which is to say, what we most want to hear.