Carnal Reason

Equality, considered as a political ideal, originally meant that everyone should be subject to the same set of laws. This was a radical idea, denying the legitimacy of the legal advantages that noblemen and clergy once enjoyed over the common man. The modern idea of equality is that everyone should enjoy the same success, regardless of personal ability or effort. The new equality can be used to extort money via the the Equality Game. It is played like this:

The are two classes of players, the protected groups and the deep pockets. Any member of a protected group can make the first move by filing a lawsuit against any deep pocket on the grounds of discriminatory hiring, firing, admissions, promotions, or whatever practice seems likely to work. The evidence of discriminatory practice can be completely statistical, something like women make up 51% of the working population, but only 43% of BigMoneyCorp employees are women.

The protected group invokes the Equal Outcomes Axiom:

Given any deep pocket D, and any protected group G,
Probability(G|D) < Probability(G) implies that D is discriminating against G.

Probability(G|D) is the ratio (# members of G who are also in D)/(# members of D).
Probability(G) is the ratio (# members of G)/(total population).

To make matters concrete, consider the situation at D=BigMoneyCorp with G=Women.
Probability(G|D) = .43 < Probability(G) = .51, so discrimination has been demonstrated.

The mathematical implications of the axiom are interesting.

Suppose you run a deep pocket organization D, and you want to avoid discrimination in hiring. You cannot allow any group G to be under-represented in your organization. Neither can you allow any group to be over-represented, since an over-representation of one group must result in the under-representation of another. You must ensure that Probability(G|D) = Probability(G), for every group G.

This means that membership in D and G must be statistically independent. Statistical independence requires that D cannot take group membership into account in its hiring practices, which is well and good. While that is required, it is not sufficient. Statistical dependencies can arise if different protected groups have different preferences with respect to working at D. Suppose D is a pig farm; you may have some trouble hiring your share of Muslims. So much the worse for you.

Fairness per the Equal Outcomes Axiom requires not only non-discriminatory employers, but also non-discriminatory employees. A fair outcome is possible only if the protected groups exhibit no systematic employer preferences. The fact that such employees do not exist in appreciable numbers should be seen as a feature, not a bug. It ensures that the EOA will continue to magically squeeze money out of deep pockets.

If you are playing the Equality Game as a protected group G, here's a tip. Recall that the EOA requires the deep pocket D to prove that
(# members of G who are also in D)/(# members of D) >= (# members of G)/(total population).

There is some ambiguity in the definition of the total population which you may be able to exploit. Find a deep pocket which operates in multiple legal jurisdictions. With any luck, you may find (for example) that while G is well represented in D at the national level, some particular state has a higher density of G than the national average, and G is under-represented in D within that state. Creative shopping will often find an advantageous venue.

Posted on June 1st, 2005 by pwyll
Filed under: General, law, politics