> No, you cannot expect overlapping, non-mutually-exclusive sub-populations of 9% and 18% to equal 27% of a total population.
Obviously, this is true. I don't know why you continue to exclude the Latino population, but again Latino CS grads (to focus on that narrow slice of tech jobs) are also ~7%.
> I hope that helps explain why your expectation of the gender & racial diversity of CS graduates is inaccurate.
But (as you have pointed out) we do know more about the sub-populations that let us infer that the populations are less overlapping than we might prefer. If you're interested, there are published demographic data that go into more detail than I have time to do here. But yes, I understand the concept that a Black Latina fits into three categories.
> you'll extrapolate (incorrectly) that 2/3rds of this population meets your criteria.
The basic idea here is simple: at Microsoft scale, their pipelines broadly should look like the collegiate exit pipeline. If you think that < 30% of programmers are women and/or Black/Latino, I think that assumption is the problem.
No matter what assumptions you make around the numbers, my core point still holds. Which is that if your pipeline is routinely missing X% of qualified applicants, your sourcing is not good enough to know whether you're even seeing the best candidates. That is as true if X% is 15% or 30%.
Using an analogy: if your company isn't seeing any applicants from states representing 15% of the population (say: Texas and New York), that says more about your sourcing than the talent available.
Obviously, this is true. I don't know why you continue to exclude the Latino population, but again Latino CS grads (to focus on that narrow slice of tech jobs) are also ~7%.
> I hope that helps explain why your expectation of the gender & racial diversity of CS graduates is inaccurate.
But (as you have pointed out) we do know more about the sub-populations that let us infer that the populations are less overlapping than we might prefer. If you're interested, there are published demographic data that go into more detail than I have time to do here. But yes, I understand the concept that a Black Latina fits into three categories.
> you'll extrapolate (incorrectly) that 2/3rds of this population meets your criteria.
The basic idea here is simple: at Microsoft scale, their pipelines broadly should look like the collegiate exit pipeline. If you think that < 30% of programmers are women and/or Black/Latino, I think that assumption is the problem.
No matter what assumptions you make around the numbers, my core point still holds. Which is that if your pipeline is routinely missing X% of qualified applicants, your sourcing is not good enough to know whether you're even seeing the best candidates. That is as true if X% is 15% or 30%.
Using an analogy: if your company isn't seeing any applicants from states representing 15% of the population (say: Texas and New York), that says more about your sourcing than the talent available.