You cannot forcibly hire someone who would prefer to work elsewhere (if that someone can really work elsewhere, e.g. not be worried with matters of survival and

You cannot forcibly hire someone who would prefer to work elsewhere (if that someone can really work elsewhere, e.g. not be worried with matters of survival and quality of life). This is probably a reason that in a lot of developing countries (e.g. India) gender gap in IT is much smaller than in developed ones (e.g. Norway, #1 gender equal country)- salary is choosen over comfort. Now, imagine a situation: 100 persons (80 men, 20 women - ratio observed by actual HR) are looking for a job in Google, which has 10 open positions for them. Google wants to hire the best possible candidate - and hires 8 men and 2 women; both genders are choosed from top 10% of candidates of a same gender (Assuming their skill does not correlate with their gender). That hiring policy does not reduces observed gender gap between candidates at all. So let's introduce positive discrimination (so called affirmative action), and say, that Google wants to hire 5 men and 5 women (1:1 ratio, but even 7:3 will suffice to show worsening effect, albeit in weaker proportions). Now men are choosed from top 6.25%, and women - just from top 25%. The is no gender gap anymore, but now there is a considerable skill gap - top 6.25% men are having considerable advantage over 25% women. More to that: men are not happy because they think they have struggled much more for the same position than women; women are not happy because they think they were hired not for their skills, but for their gender. Also,both men and women being paid based on their performance alone will result if considerable wage gap. The original memo stated the same, albeit in less transparent form: personal skills cannot be measured with 100% accuracy, so there is always a chance to hire underperforming person (false-positive) or not to hire well performing person (false-negative). Positive discrimination radically reduces false-negatives, which inevitabely magnifies number of false-positives (see Type I and Type II errors in statistics and Neyman-Pearson criterion). Author proposed not to fight on-enter gender gap with discrimination, which would result in negative effect for company's performance, but to make company's environment to be a more attractive (which may reduce on-enter gender gap, reduce number or women leaving industry permanently and won't compromise performance)