Hacker News new | ask | show | jobs
by Mithaldu 3875 days ago
HN moderators: It would probably be best if you buried my post.

I idiotically misread the chart posted in the post. I originally thought the chart under ( https://cdn-images-1.medium.com/max/800/1*F9MLGQLTU2HZD4In6o... ) wasn't normalized to anything; but upon rereading i found that the top row is the comparison row and represents the percentages of working force age us citizens in the tech industry, which Twitter with its 1% falls far below.

I'm sorry for my mistake.

My original post below for context for the replies made to it, but which is otherwise useless.

------------------------------

It's alway a little confusing when people bring up diversity reports that aren't normalized to an appropiate comparison metric. (Possibly local demographics, or any number of more in-depth metrics. I also earlier suggested applicant demographics, but justizin pointed out those are not feasible.)

Also see: https://xkcd.com/1138/

To bring it into contrast with the article, he says:

"<5% make up engineering and product management combined."

According to this census report: http://www.bayareacensus.ca.gov/counties/SanFranciscoCounty....

Race - Black or African American 48,870 6.1%

The twitter numbers are thus a little below average, but not necessarily unexpected given where Twitter's headquarter is located.

Edit:

According to this, Twitter has roughly 30% asians: https://blog.twitter.com/2014/building-a-twitter-we-can-be-p...

Which seems to fit the census as well:

Race - Asian 267,915 33.3%

Another Edit:

To clarify, i am not saying that there is no problem. I'm merely saying that in order to solve a problem, one must both set appropiate goals, as well as correctly identify the root cause. Both of these can only be done usefully by applying statics correctly.

6 comments

Based on my own cursory observations of the population flux in the Bay Area, I believe your analysis and thus conclusion is based on a false premise.

That analysis assumes that the hiring base is predominately drawn from local people. Speaking anecdotally, I know many people who graduated from college and headed to the Bay Area to look for a job. I myself moved to Mountain View for a job, then later moved away.

A more complete analysis along the lines you took would look at where people in engineering and product management were, say, 1 or 5 years previous. That are would likely better characterize the relevant population statistics.

As an obviously contrived example, suppose people in engineering and product management are only in the Bay Area for 4 years, burn out, and leave, and suppose the companies offer free relocation from anywhere in the US. Then it doesn't make sense to look at the local demographics.

As a more real-world case, consider a place like Los Alamos National Labs, which has a large number of people from around the world working there, with a relatively high turnover rate partially due to interns, post-docs, and visiting professors. Los Alamos county was carved out for the lab, so the demographics of the county reflect the lab, but most of the people working at the lab are not from Los Alamos.

I do not have access to these sorts of numbers, merely pointing out how it's not easy to interpret the numbers you gave, or the certainty of your conclusion that it's a "not unexpected" result. I suspect the uncertainty is actually very high, and a shot-in-the-dark/back-of-the-envelope estimate isn't likely to be useful.

I did express the thought of "normalized to an appropiate" metric badly. I don't know what that metric is, and don't presume to know. Local demographics were just one example to me. I'll correct the post.
I quoted your use of "not unexpected" because it suggests that you have some expectation. I would call the the basis for that expectation part of a presumption that you know.
Yeah, i can see how it could be interpreted like that. I didn't have any strong expectation of how it should be. Merely a number of guesses as where that number could end up. To me it seems reasonable that the number would end up somewhere between half of the local percentage and 1.5x of the us percentage (assuming foreigners are not statistically significant), which would put it somewhere between 2.5% and 25%. To me numbers outside of that would be unexpected and surprising.

Interestingly, the 30% number seems to indicate that twitter actually does pretty good user-wise.

Your own statistics reference indicates that 36% of the Bay Area population is foreign born, and 16% do not have US citizenship, so you cannot assume that "foreigners are not statistically significant."
Why not look at the areas where people in engineering and product management came from? It takes very few steps before we start expecting all metrics to resemble global metrics.
Does Twitter hire only from the local area? If they have a relocation program, the local demographics are a lot less relevant.
Very fair point, and a little hard to tackle. I have no solid answer either as to how to get fully correct statistics that actually help in evaluating just how bad the situation is, and where to start working on it. But, as you point out: One must take into account many factors in detail in order to get anything related to reality.
There are also many factors that have no bearing on a given matter. Bringing up local statistics in this instance can be interpreted as deliberate equivocation, which could explain some of the strong reactions to your comment.
Possible.

Personally i was thinking along the lines of: Maybe the real problem is whichever property of the area they are headquartered in prevents higher amounts of black people from living there. Which may, among others, turn out to be: Companies there are biased against black people, so they're relocated less often. However without looking at the situation closely, any conclusions drawn are premature.

Without fully correct statistics, how can anyone objectively say that there is a problem in the first place?
Twitter definitely has a relocation program. And in general, engineers move to the Bay Area in order to work at companies like Twitter.
Alameda county right across the bridge is 12.6% black or african american, and Oakland in particular is 28%. Of course I think all this bears little relevance to a company that recruits candidates from pretty much anywhere.
Most bay area tech workers have moved here after getting a job so the local census doesn't mean much.
It's even actually reenforcing itself based on the bias
That's a solid point. I don't know that that data is so accessible to anyone outside of a given companies HR department.
Your numbers break down for the Hispanic\Latino population according to your own sources. The 2010 census says San Francisco has a 15.1% Hispanic population and Twitter has 3% Hispanic employees.
They only break down if you assume i'm saying there is no problem. ;)