You often see the argument that the central political divide is social class, not race and ethnicity. If this were true, then we should see it in voting patterns.
The General Social Survey asked participants who they voted for in 2016, and they also asked about annual income and race. The results below are estimates from a logistic regression model that predicts 2016 voting for President with income and race. The three racial categories are white, black, and other race. Whites are the reference category and so are omitted from the model (sample size = 1,360).
From the p-values (probability) you can see that, once you adjust for race, REALINC (inflation-adjusted income) does not significantly predict who you voted for in 2016. By contrast, blacks and people of other non-white races were less likely than whites to vote for Trump, regardless of one's social class.
Race trumps.