Sadly, more bad "science" coming out of social psychology. In the first study with the political speeches, the "competence" words were words taken from lists of words used to convey one's own "status", "dominance", and level in the "social hierarchy". Lower scores did not imply the candidate tried to make themselves look incompetent, just that they did not make and effort to emphasize their own social dominance.
Plus, the audiences that Repub and Dem candidates would have spoken to would differ in important ways. The Dems spoke to almost twice as many mostly-minority audiences as the Republicans did. The audiences were only counted as "mostly-minority" if there was something about the location or the group that allowed the researchers to assume it was mostly nonwhite, such as a speech about Cuban Independence Day or to a mostly minority rural town, etc.. Candidates speak mostly to groups they think they have a shot at winning, which would inherently mean the Dem and Republicans are speaking to different subsets of minorities. Dems would be more likely to speak to just a general audience of black Americans where education levels are low. Repubs are more likely to either speak to very select minority groups that lean Republicans b/c they are business owners (who are also more educated) or Cubans who lean republican and are twice as likely to have a college degree than Hispanics in the US overall.
Besides, Republicans showed almost the same difference in how they spoke to white vs. minority audiences. The difference in their score was 55 points vs. 63 for the Democrats. The only reason the difference did not reach statistical significance for the Republicans is because their sample size was much smaller, due to only giving 14 speeches to minority audiences in 25 years that the study looked at.
As for the other studies using samples of college students, they do not actually have the people interacting with real people and do not actually manipulate the race of the fictional interaction partner. Instead, they manipulate the name given to this fictional persons as either "Emily" or "Lakisha". IOW, the name designed to indicate the person is "black" was a highly uncommon name that has never been among the top 1000 names for US born babies, whereas Emily is the #1 most common name among current college age people.
That means Lakisha is not even common among American blacks, and there is a high probability it is used mostly by mothers who are lower in education relative to average blacks. Plus, the name is Swahili in origin and thus hits the English speaking ear as being foreign.
If you are speaking to a person who is plausibly a non-native English speaker wouldn't it make sense to use "happy" rather than "euphoric"? And yes, that is the choice research subjects were given. They were given specific words to choose between, and all the words categorized as "competent" were less common words less likely to be known by someone with less education or who speak English as a second language. Note that the research effects were tiny, often non-significant and inconsistent, and the samples were small. So, if even a couple of liberals interpreted "Lakisha" as possibly being a non-native English speaker, this would easily account for the results. So, what it likely means is that liberals are more likely to consider the language skills of their audience and try to avoid confusion just for the sake of appearing smart. What it definitely does not tell us is anything about how liberals view or speak to "black people", since the name "Lakisha" is atypical for black people and indicates things other than race.