Some interesting notes on polls

So the big takeaway from tonight was how inaccurate the polls were in Michigan. As you would expect, the polls have been off to varying degrees in every state so far, but what is interesting is the pattern that has emerged. The numbers below is how much the results differed from the polls, and who benefited (so "Clinton +1" means Hillary did 1% better than the polls predicted). For polls I'm using the aggregate numbers from fivethirtyeight, since individual poll numbers vary so widely:

Clinton wins
Iowa (Sanders +5)
Nevada (Clinton +1)
South Carolina (Clinton +12)
Alabama (Clinton +12)
Arkansas (Clinton +10)
Georgia (Clinton +4)
Massachusetts (Sanders +7)
Tennessee (Clinton +7)
Texas (Clinton +1)
Virginia (Clinton +4)
Louisiana (Sanders +8)
Mississippi (Clinton +2)

Sanders wins
New Hampshire (Sanders +8)
Colorado (Sanders +12)
Minnesota (Sanders +26)
Oklahoma (Sanders +11)
Vermont (Clinton +6)
Maine (Sanders +14)
Michigan (Sanders +24)

So in the majority of cases, in the states where each candidate won, they also did better than the polls predicted. The exceptions were Iowa, Massachusetts, Louisiana, and (somewhat surprisingly) Vermont.

It's also important to note that Bernie had a couple of HUGE surprises, Minnesota and Michigan, which have demographics somewhat similar to two states voting next week (Illinois and Ohio). Hillary is currently way ahead in the polls in both states, but I won't be surprised at all to see Bernie wins there.