Just a quick comment to point out that all the analysis of how Democrats or Republicans gained or lost among populations in the last election is largely bunk.
In general, it is extremely difficult to accurately compare different elections because you are not looking at the same sample group. So, right off the bat, if you see any comparisons between the party support among demographic groups in 2012 and 2014, ignore them. Apples to oranges. Comparing 2010 to 2014 is better, but still problematic. Consider: Republican support among Latinos went up from 2010 to 2014. So, did some Latinos who supported Democrats in 2010 switch to Republicans in 2014, or did those who supported Democrats in 2010 simply not vote in 2014? Both cases would show up as rising support among this population.
Now consider that turnout for 2014 was the lowest since World War 2. This suggests the second of the possibilities—Democratic voters staying home—is more likely: Republican voters were simply more committed, and so those Republicans in all groups were a larger percentage of the total electorate. So what really happened was not that Republicans gained among Latinos, but that Democrats weren’t motivated enough to vote. In fact, this is largely what happened (in reverse) in 2006, when Republicans stayed home. Regardless, this is a somewhat different issue, and should not be confused with fundamental changes in support. It’s a fundamental change in turnout, likely stemming from a fundamental change in motivation, and while that deserves study, we have to be certain we’re studying the right thing.
(And by the way, I would suggest that motivation comes from having a well-defined enemy more than a well-defined hero, so the out-party will always have an advantage there.)
UPDATE: fivethirtyeight.com weighs in, making a similar point. When turnout is high only in Republican states, overall turnout will look very Republican. A turnout problem, then, not a change in support among key demographics.