Statistically significant, such that it created some academic literature and enough cover for Harvard to make this new policy. Practically significant, maybe not.
Here's a comment from a different branch of this thread:
> The findings in general is that much of the disparity between minorities and whites goes away for harder analogies that are more technical or scientific in nature, in fact contrary to what I expected, the recommendation of the study is to simply eliminate the bottom third of verbal questions (ranked by difficulty) involving analogies or sentence completion, as those are the questions that are most likely to involve cultural bias.
First off, the SAT has gone through many revisions in the last 20 years so it's unclear this still applies. The SAT doesn't even have analogies any more, perhaps as a result of this research.
Secondly, that doesn't address differential prediction abilities. It addresses a concerning aspect of the verbal test, but is somewhat irrelevant when your goal is to predict college success. I'm not aware of differences on that metric.
Here's a comment from a different branch of this thread:
> The findings in general is that much of the disparity between minorities and whites goes away for harder analogies that are more technical or scientific in nature, in fact contrary to what I expected, the recommendation of the study is to simply eliminate the bottom third of verbal questions (ranked by difficulty) involving analogies or sentence completion, as those are the questions that are most likely to involve cultural bias.
https://news.ycombinator.com/item?id=29599455#29600100