I'm dubious about the whole survey. The data is online. The survey was conducted by an academic consortium. The survey was divided up into 50 segments. There were 120 common questions, and 120 questions that a researcher could ask that were specific to their research. They could also choose which persons were in their segment. So a researcher might pay for a 1200-person national sample of old people or millennials, etc.
But the survey was conducted on line using Yougov. Respondents were self-selecting. They had to use the Internet and volunteer to be surveyed. Even if you do demographic matching so you get the right percentage of various demographic groups, you still have that selection bias.
Based on the raw numbers it appears that President Clinton was elected by a 48% to 41% margin, with 11% voting for someone else.
Also, lets say that 50% of respondents were Democrats, and 25% supported Sanders, what percentage actually voted in a primary? 10% of the total (i.e. 40% primary turnout). So that would mean that 1% of the total were Trump-Sanders supporters. But relative sampling error goes up as the number of respondents goes down.
