Hacker News new | ask | show | jobs
by rayiner 229 days ago
These studies are misleading, because they try to create race signals by using names that are also class signals: https://newsroom.ucla.edu/releases/ucla-study-suggests-resea....

Also, the study suggests that, even with this flawed methodology, a bulk of industries are in the least discrimination category with only a 3% lower callback rate for “black sounding names.”
1 comments
Dylan16807 229 days ago
Do you have any numbers for trying to correct for that factor?

And the bulk are not at 3%, the bulk are between 5 and 10. 3 was the absolute lowest.

Also you didn't mention the CEO thing, does that mean my numbers were sufficient to address that worry?

link
rayiner 228 days ago
As I read it, the industries were grouped into three categories. “Least discriminatory” was at 3%. Those are all the industries in green. These are small differences in a study that’s not well designed to begin with.

The explicit discrimination in universities against whites and asians is huge in comparison: https://nypost.com/2023/06/29/supreme-court-affirmative-acti.... A black applicant to Harvard with an academic index in the 5th decile had an 800-900% higher chance of admission than a white or asian candidate with the same qualifications. This isn’t just CEOs. The pattern was similar at UNC, a state school.

link
Dylan16807 227 days ago
> As I read it, the industries were grouped into three categories. “Least discriminatory” was at 3%.

The least single industry was 3%. And each single industry is a very noisy data point, based on a couple companies and needing more data points. By the time you aggregate into more solid data, like those bigger categories, it's more than 3%.

But the whole thing could use better methods and more data for sure.

> The explicit discrimination in universities against whites and asians is huge in comparison

In comparison to this specific resume effect it's pretty big, but that was just a basic example, not an attempt to list the biggest issue.

In comparison to the fortune 500 CEOs the overall effect here is smaller (no I'm not going to look at 5th decile in isolation).

Also even after this bias was applied, they're admitting a below-population-average amount of black students and a far above-population-average amount of asian students. So there's a bunch of other data necessary to properly analyze what's going on and how bad it is. Should there be a super tight correlation to academic decile? There are huge differences in school quality that muddy the signals, and those differences often correlate with race.

I'm not saying they did nothing wrong, but I'm saying it's unclear what the numbers should have been.

link