I have a different take on this. The following are all possible:
... are less likely to earn ...
... are less likely to have earned ...
... were less likely to earn ...
... were less likely to have earned ...
If the sentence is taken from research that studied a certain group of African Americans and Whites, the researchers do not know what the future of the population will be. They can only report the past:
The African Americans were less likely to have earned a bachelor's degree than their peers.
Note: "The ..." indicates a specific cohort, or group studied.
From this, they might make a more general claim:
African Americans were less likely to earn a bachelor's degree than their peers. (In the time period related to the study. The absence of "The ..." means all African Americans, or African Americans in general.)
From the result of this study, they might predict the future generally:
African Americans are less likely to earn a bachelor's degree than their peers. (All)
Then, if they have a new matched group in front of them, they might predict:
The African Americans are less likely to have earned a bachelor's degree than their peers.
They can also predict:
"In ten years from now, we expect to find that the African Americans studied will be less likely to have earned .... This would indicate, if statistically significant, that African Americans in general will have been less likely to have earned ... "
etc.