I guess you think the groupings on the chart are really worthwhile and well-presented?
Because I sure don't.
I have to say I found the graphs pretty easy to understand, although clearly I was the exception and so would not claim the graphs were well-designed. I do think it is true that the graphs don't give any sense of the relative sizes of the cohorts, although I don't actually think varying the bar area and/or vertical size would help people understand it more clearly.
In terms of being worthwhile, uh, of course they are. If you don't think it is worthwhile to analyze voter behavior in the context of previous votes I can't really imagine what you would find worthwhile. It seems like literally the most important piece of data we could get!
But regardless of your view on that it should be pretty obvious that literally eliminating the entire study by reducing it to a poll and taking out every piece of data is not a way to understand the data. It's a way to discard the data.
What's up with those charts?
How do they make sense
The goal of the study is to look at six separate cohorts, based on who they voted for in 2012 (Obama or Romney) and what they did in 2016 (Hillary, Trump, or sit out). It should be fairly clear that if Hillary just won every vote Obama won in 2012 and lost every vote Obama lost in 2012 she would have won. So the change in voter behavior is quite relevant to understanding the election.
Obama-Clinton and Romney-Trump voters are fairly basic and presented mainly as contrast -- these are reliable voters on either side. Obama-Other and Obama-Trump voters are the fabled lost votes that Hillary needed in places like Michigan. Romney-Other and Romney-Clinton, by contrast, are the hypothetical NeverTrump voters whom Hillary worked hard to sever from the GOP.
As you can see, Obama-Other voters look very similar to Obama-Clinton voters. These voters aren't pleased with what their lassitude has wrought. They voted in 2018 for the Democrats and they mostly intend to vote for the Democrat in 2020. So we can understand these voters probably as fairly usual Democratic voters who had a particular distaste for Hillary and who would like to return to the fold.
Romney-Clinton voters, who voted GOP but then refused to vote for Trump, have an even higher disapproval rating of Trump, and in 2018 extended that distaste to voting out Republicans. These are the voters who flipped Orange County, and they intend to vote again against Trump in 2020. Trump has transformed this bloc of voters into a bloc that looks nearly identical to reliable Democrats.
On the other side, Romney-Other voters, the McMullin contingent, are more confused. They do not much like Trump, but they voted GOP in 2018, presumably because they love the party but not the man. They are fairly evenly split on Trump in 2020 -- they will not reliably return to support him. This is a potential weak spot for Trump, who would like to pick them back up.
The last interesting group is Obama-Trump. In general, as Trump voters, these guys are fairly satisfied with Trump. But, as former Democrats, they crossed over in 2018 more than any other cohort, and a sizeable fraction of them intend to return in 2020. A lot of these voters are gone forever, but a significant percentage are wavering.
Overall, as the chart titles suggest, Republican-leaning cohort are simply more torn, and it's not purely hypothetical -- they put action to word in 2018 when we won our House wave. We know they are genuinely willing to vote Democratic. Meanwhile the reliable Democrats are fairly confident that they will turn out against Trump. If these patterns play out as suggested, it's very bad news for Trump.
www.voterstudygroup.org
