Does Clinton have a ceiling?

The story of the week on the Democratic side this week was Bernie Sanders’ impressive win in Michigan, where polls had him trailing 20%.  It’s probably right to question the demographic models pollsters are using for the upper-Midwest states (including Ohio and Illinois that vote March 15 – and where Clinton holds 20-point leads).

This blog isn’t going to do that. Our model is set up to assume that the candidates’ polling averages are their “floor”.  In Michigan, for example, polls showed Clinton at around 58% support and Sanders at 38%.  Our model split the remaining votes proportionally (i.e. Clinton got 58% of undecideds).

In the event, Clinton’s strength was oversold by around 10%, and Sanders, in effect, won all of those plus all of the undecideds.  Since polls are the only data that show likely support, changing that as a baseline for the model would put us too far into guess territory.  Instead, we’ve built into the model a couple of variables to re-align how undecided votes are split.

National polls have Clinton leading Sanders 51.8% to 37.6%. So what happens if all of the undecideds vote for Sanders (i.e. if he is polling 48.2%)?

Well, the race closer, and ends with a different result from our baseline projection. Clinton would end up 150 delegates short of the 2,382 delegates the Democrats need to win on the first ballot without superdelegates. Most outlets show her with around 460 superdelegates pledged, so she would likely still win the nomination, but with Sanders saying he’s in until the convention, the backroom politicking would continue until the actual nomination.

How likely is this outcome? The polls suggest Clinton is holding/gaining nationally and in most states where there is polling.  Obviously, that’s not how things played out in Michigan.  The key states to watch for assessing both whether the polls are overplaying Clinton’s strength and whether late deciders are breaking to Sanders are Florida, Illinois, and North Carolina, and Ohio (there just isn’t good polling in Missouri).  Until then, the current polls are the best information we have.