If you were to accept the current estimates on IQ, where blacks have an average IQ of 85, Hispanics an average IQ of 90, whites an average IQ of 100 and Asians an average IQ of 105 with a standard deviation of 15 by design, and that Youtube only hires people two standard deviations from the mean, then about 2% of employees would be black, 5% Hispanic, 30% white and 63% Asian. I am quite sure that whites are over represented at the expense of Asians.

This assumes that there are equal number of people in the hiring pool in each category. There is nothing to suggest that the applicants that apply are drawn at random from the population.

Also the absolute numbers of each groups also matter. On this basis we should expect far few Asian and Black employees just on the basis that there are fewer than whites within the USA [0].

0. This is assuming the employee pool is drawn from the USA population.

Ok. California is 39% Hispanic, 38.8% White, 5.8% Black and 13% Asian. California has 40 Million people. Let's say half have not graduated college yet and are not in the employee pool. That means 8 Million Hispanics, 8 Million whites, 1.16 Million Blacks and 2.6 Million Asians. Taking this new information into account, Asians should be around 37% of the employees, Blacks less than 1%, Hispanics around 9% and Whites 54%. Google's numbers say 61% White, 30% Asian, 2% Black and 3% Hispanic. So, it's both Hispanics and Asians who are under represented.

Based on those assumptions, which I'll iterate below first, let's get the numbers:

1) hiring based on IQ, cutoff at 2 standard deviations above the global mean (mu = 100, sigma = 15, by design), 2 sigma above that makes 130. You get hired if you're the candidate with the highest IQ, if you satisfy the minimum of an IQ of 130

2) 8/20 hispanic, 8/20 white, 1.16/20 black, 2.6/20 asians (and let's just pretned that sums to 100%). Or: 40% hispanic, 40% white, 5.8% black, 13% asian

3) let's assume 1000 candidates for each position.

So each round has 1000 candidates: 400 hispanics, IQ taken from N(90, 15) 400 whites, IQ taken from N(100, 15) 58 black, IQ taken from N(85, 15) 130 asians, IQ taken from N(105, 15)

The numbers: 33.42% Asians, 10.85% Hispanic, 0.54% Black, 55.20% White

Odds of getting hired under those criteria: 0.25% Asians, 0.02% Hispanics, 0.01% Black, 0.13% White

And that's why nobody's going to be happy with expected outcomes. Just imagine the (completely "fair") news headline "Asians 25 TIMES more likely to get hired than blacks in the bay area".

  import random

  counts = {'h': 0, 'w': 0, 'b': 0, 'a': 0, None: 0}
  experiments = 10000

  for x in range(experiments):
    candidates = []
    for h in range(400):
      iq = random.normalvariate(90, 15)
      candidates.append((iq, 'h'))
    for w in range(400):
      iq = random.normalvariate(100, 15)
      candidates.append((iq, 'w'))
    for b in range(58):
      iq = random.normalvariate(85, 15)
      candidates.append((iq, 'b'))
    for a in range(130):
      iq = random.normalvariate(105, 15)
      candidates.append((iq, 'a'))

    # filter iq > 130
    candidates = [(iq, typ) for (iq, typ) in candidates if iq > 130]

    if candidates:
      selected = sorted(candidates, key=lambda (x,y):x)[0][1]
    else:
      selected = None
    
    counts[selected] += 1
       
  total = sum(counts.values())

  print total
  for k, v in counts.items():
    print "%s %2.2f" % (k, 100.0 * v/total)
  
  print "odds of hire if hispanic : %2.4f%%" % (100.0*counts['h']/experiments / 400)
  print "odds of hire if white : %2.4f%%" % (100.0*counts['w']/experiments / 400)
  print "odds of hire if black : %2.4f%%" % (100.0*counts['b']/experiments / 58)
  print "odds of hire if asian : %2.4f%%" % (100.0*counts['a']/experiments / 130)
  print "odds of no hire at all: %2.4f%%" % (100.0*counts[None]/experiments)

> Odds of getting hired under those criteria: 0.25% Asians, 0.02% Hispanics, 0.01% Black, 0.13% White

> about 2% of employees would be black, 5% Hispanic, 30% white and 63% Asian.

These two comments say the same thing. Mine is standardized.

Your take away from all this is a little strange. The Google numbers say that Asians are 25 times more likely to be hired than blacks. That has happened in real life. No one complains about that. I have read that Google is too Asian though.

People harp about less-talented people not deserving to be there, but that is already the case. From a statistical point of view, Whites are over represented at the expense of Hispanics and Asians with a higher IQ. Real life is messy, but you get the idea.

This battle has been fought already and enough people at Google were convinced that there was a bias in their hiring practices. So, now they are in the process of correcting.

You are going to have trouble fitting all of this into a tweet.

It has math and numbers. Nobody interested in that on Twitter anyway.

What does the first number in N(90, 15) mean? the median IQ of a given race? I guess the truth is hurting people's feeling.

It's shorthand for a normal distribution with mu=90 sigma=15.

What percentage of those populations have the requisite secondary education and or experience (coding, statistics, marketing, electrical engineering) to work at Google? Before you try to just match up a general population to a workforce let's at least try to have a representative sample of what the "pool of qualified applicants" looks like.

I assume you are saying that Blacks and Hispanics will not meet the educational standards. Asians will and they are under represented.

The level of the problem doesn't matter. Working at Google is an elite job. We could go a level lower to Stanford. Asians are under represented at Stanford. We could go down a level to an elite California High School. Asians are still under represented. Blacks and Hispanics are really under represented at this level. It only gets worse for them the lower you go at elite institutions.

I use the general population because that is where the workers are coming from. If the pool of qualified applicants does not match the general population, then somewhere the qualification process failed.

> If the pool of qualified applicants does not match the general population, then somewhere the qualification process failed.

That's an interesting conclusion, statistically and as a societal commentary. Assuming for a second we take your conclusion as given, and we look at the legal liability question that someone in alphabet legal or HR has to be thinking about when the make hiring policy decisions: which population should their workforce be representative of: the base population, or the population of qualified applicants?

Edit: >I assume you are saying that Blacks and Hispanics will not meet the educational standards. Asians will and they are under represented.

I don't care I'm here for the legalities and statistics discussions, see my other comments here about judging people based on melanin or Genetics. It's silly where it isn't insulting to reduce a real person based on their inclusion/exclusion to poorly defined groups with little to no impact on actual ability to perform/qualify.

The guy asked a question, can we stop making these assumptions? It's childish, it's more constructive to provide the statistics the inquirer asked for.

How are you drawing the conclusion that IQ is correlated with hiring value?

You’re correct that the difference is in the margins. But, at the end of the day, is any exclusion based on race acceptable? From my perspective, it’s hypocritical to champion equal opportunity while systematically having a different hiring process for different races, not to mention illegal.

That is not the argument I am making. I am asking the question if there is a super majority of white and asian males being hired at these firms (that is present in the Google stats), is it fair to say on an individual basis that the reason you did not get the job was because you were passed over for someone unqualified in an underrepresented group, or is it because said person just did not get hired for some other reason?

I think you have put your finger on why these programs are bad for everyone. They really help nobody in the long run and create enormous blowback.

A pretty straightforward explanation is that ⅔ [1] of the tech workers in Silicon Valley are immigrants. Silicon Valley recruits outliers from around the world, and it doesn't make sense to apply US demographics to an industry that recruits from such a pool.

If you assume that the Silicon Valley percent of immigrant workers holds for Google you conclude that 6% of the American Googlers are African Americans, since none of the foreigners can be. About 2.4% [2] of computer science degree recipients in the US are African American.

[1] https://jointventure.org/images/stories/pdf/index2018.pdf

[2] https://cra.org/crn/2015/09/expanding-the-pipeline-the-state...