This is wrong. By this logic, "25% of people have degrees and 100% of degree holders have parents with degrees" is better than the 4%/20%, because it's only 4 t

This is wrong. By this logic, "25% of people have degrees and 100% of degree holders have parents with degrees" is better than the 4%/20%, because it's only 4 times instead of 5 times. One of the better methods would be to calculate the "Bayes factor"(https://en.wikipedia.org/wiki/Bayes_factor https://en.wikipedia.org/wiki/Bayes_factor) for estimating possibility of holding degree given the evidence of "parents holding degree". In this case, we need to know the parents degree percentage for both people in 1970 and 2020.