Pat and I both ask the same question on our Manifold Survey:

Treating these as four categories for the purpose of a diversity index calculation, which of our surveys will have the highest 

 of order 2? This is equal to the reciprocal of the Simpson index, which is the weighted arithmetic mean of the proportion of respondents who select each option, with the weights being the proportions themselves. 

This is also equivalent to asking for which survey the probability that two random respondents have different genders is higher. 

The survey is officially out! You can take it here: 