That's the question Tom Edsall asks in his latest New York Times column. Well, the answer to that is pretty simple: they aren't black and black voters are so overwhelmingly loyal to the Democrats for very specific and well-known historical reasons.
So, not a big mystery there. That said, Edsall seems to be implying that Democrats are significantly underperforming among this group relative to what one would reasonably expect from how awful Trump is, etc.
I'm not so sure about that. The best data we have on Latino support rates from Catalist does indicate that the Democrats did very well indeed among this group in both 2016 and 2018--significantly better than 2012 and especially 2014. Catalist says Clinton carried Hispanics 71-24 (+47), compared to Obama's 67-30 (+37) in 2012 , and that House Democrats carried the group 71-27 (+44) in 2018.
So they may not be an 80-20 group but in the current environment but they do look like a 70-25 group, which is still pretty darn good. You can see this rough pattern in a number of other states where Catalist data are available like Arizona, Georgia, Iowa, Minnesota and Virginia, where Democratic margins were generally in the +40-+50 range. Of course, there were states where Catalist data are not available like Florida where the margins were presumably lower, as was also likely the case in various noncompetitive races in other states.
But the central tendency of this group is very strong and should not be underestimated. 70-25 is a heck of baseline to start with even if you're not guaranteed to get that in every election in every state.
Should we expect this baseline to continue to ratchet up toward 80-20, say, if Trump and the GOP continue on their current course? I am doubtful. Hispanics are motivated by many other issues besides immigration, some are conservative and will remain so, some are evangelical Protestants and so on. In that sense, I think Edsall is right that Democrats who are relying explicitly or implicitly on this group becoming as monolithically Democratic as blacks will wind up disappointed.
I think the bigger problem with Latinos for Democrats lies not in their support rates at this point, but in their relatively poor turnout. This problem is well-documented and conceivably could be at least partially solved by good old-fashioned mobilization efforts. I'd worry about that rather than why Latinos don't vote 80-20 Democratic.
Finally, as I've noted a number of times, Latinos by themselves are not the solution for Democrats even in Latino-heavy states like Arizona and Texas. Swings in the white vote, including both college and noncollege, have to be joined with strong performance among Hispanics to carry these states in 2020.