Going into Tuesday, everyone figured Hillary Clinton would win in Michigan. Both campaigns appeared to think so, though both also insisted that the race was closer than polling indicated. The Real Clear Politics average of polls in the state figured that Clinton had a 20-point advantage; FiveThirtyEight gave her a 99 percent chance of winning.
And then Bernie Sanders won.
It's not clear why he won yet, though his overperformance with black voters was certainly one reason. There was only one poll from a major pollster in the last week, but it showed Clinton with a wide lead. FiveThirtyEight noted that no one in their system polled after the Democratic debate on Sunday night. There was one poll after that showing Clinton with a wide lead, though it underestimated turnout from independents. The Huffington Post notes that turnout was higher than anticipated, which likely helped Sanders.
Setting that aside, though, it's worth noting that, while this was a big miss, it's not that unusual. Primary polling is often somewhat hit-or-miss, as most people understand. But this is also a function of the amount of attention being paid to this particular race. Had Clinton been expected to win and gone on to win by a much larger margin, had it not been primetime on a weekday, we might not have noticed.
You don't think so? Well, the polling in South Carolina was exactly that. It missed by 20 points versus the Real Clear Politics average on that Saturday, underestimating Clinton's massive win in the Palmetto State. In 2008, polls in Nevada underestimated Mitt Romney's win in Nevada by over 30 points.
Comparing results to final polling averages from 2008 and 2006, you can see the wide variance in results.
In 2008, the polls for the two parties were off between 6 and 7 points on average. This year, polling on the Republican side has been about the same, while that on the Dem side has been off by more than 8 points.
The other reason we're so focused on Michigan, of course, is that Clinton was supposed to win big but lost. Clinton should have had a big win and taken a big step toward the nomination. She didn't, and the polls didn't catch it. We very naturally don't care that Clinton's South Carolina blow-out was a mega-blowout. We do care that we didn't know who'd win a state.