That article has very suspect figures. It says that all the major candidates had supporters with median incomes above the national median income. Which tells me that the population they're polling isn't representative.
That's not cause to suspect the analysis. Yes, the sampled population may not be congruent with the nationwide population. There's a reason for that: these are exit polls at the primaries. Only those who showed up to vote in those primaries can be polled.
The VA results there, for example, also show that the majority of respondents for both Republican and Democratic primaries are over the median state (and national) _age_.
Right, that was my point - primary voters are not representative of general election voters, and so conclusions about general election voters should not be drawn from primary voter data.
The article links to its source data: http://www.cbsnews.com/elections/2016/primaries/republican/v...
The VA results there, for example, also show that the majority of respondents for both Republican and Democratic primaries are over the median state (and national) _age_.