Was that percentage equally high in 1990? Fresh grads are not qualified for management.

We can determine bias simply by comparing one company's numbers to national trends?

I said probable bias, and yes.

As you reduce geographical scope, national trends aren't as relevant or realistic.

> Because the percentage of black GitHubbers in management positions is far below e.g. the percentage of e.g. black people with CS degrees in the US

Where's the data for this? Often people talk about these things as assumptions, but I've never seen credible data. Can you share a source?

Furthermore, even the comparison that you mentioned is not accurate. It should start from a funnel like: "the number of qualified black applicants to GitHub with CS degrees", or something along those lines. GitHub can't hire people that don't apply.

Starting from only the people who apply gives a free pass to biased recruitment.

What, you want them to go kidnap black people and force them to apply?

So you don't think it's a problem if their ads say "whites only"? Or if they only advertise to white people? Or are you thinking "nobody would say whites only" and just totally unaware of the many ways that language and targeting can affect who applies to a job, sometimes skewing the pool of applicants in unintended ways? A nice, nobody-meant-it way to bias your candidate pool: only hire people who know an existing employee. White Americans mostly know white Americans.

You're asking for a citation for zero being less than non-zero? :)

He's asking for a citation that the probability of seeing zero is small (p<0.05 if you want to be frequentist about it), contingent on a proper funnel analysis and no bias.