The 80% Rule, Part I 

 I’ve blogged previously about the course in statistics and law I’m co-teaching this semester (see, for example,  here ).  The course is now in its second simulation, which deals with employment discrimination.  In a recent class, the 80% rule came up.  I wish it hadn’t.  In fact, I wish the ``rule�? had never seen the light of day.  In this post, I’ll explain what the 80% rule is.  In a subsequent post, I’ll explain why it stinks. 


 Suppose we’re interested in figuring out whether members of a protected class (say, women) are being hired, promoted, fired, disciplined, whatever at a different rate from a comparison group (say, men, and for the sake of discussion, let’s say we’re interested in hiring).  Long ago, the Equal Opportunity Employment Commission (“EEOC�?) released a statement saying that it would ordinarily regard as suspect a situation in which the hiring rate for women was less than 80% of the hiring rate for men.  Note that the EEOC has the authority to bring suit in the name of the United States against a defendant that has violated federal employment discrimination laws.  

 It would be bad enough of the EEOC used the 80% rule for the purpose it gave, i.e., a statement about how the agency would exercise its investigative and prosecutorial discretion.   Alas, courts, perhaps desperate for guidance on quantitative principles, have picked up on the idea, and some now use it as an indicator of which disparities are legally significant.  Courts do so despite the outcry of those in the quantitative community interested in such things.  More on that outcry in my next post.