What test to use to compare these two populations I have two sets of data. The first one contains the proportion of farmers that own land in 22 different towns.  The other contains the proportion of craftsmen that own land in the same 22 towns.  
I want to run a test to see if there is a difference between the proportion of farmers that own land compared to the proportion of craftsmen.  
Would Mann Whitney U test work in this case?
 A: There are several issues here.

*

*You're dealing with count proportions (count of "farmers who own land"/ count of "farmers"), so you should generally use procedures suitable for count proportions (which might in different circumstances be chi-squared tests or logistic regression or Poisson regression or loglinear models for contingency tables).


*the two proportions are not independent, since land is a limited resource. For example if lots of land is owned by farmers, there's less land available to be owned by anyone else. In some cases, you might be able to make an argument that you can reasonably choose to ignore this dependence, but that's different from pretending it's not there!
This dependence would be expected (other things being equal) to make the proportions somewhat negatively related, so a paired test wouldn't be any help in dealing with it (that is, I'm not suggesting replacing Mann-Whitney with a signed rank test, say).
What analyses you might do depends on what information you actually have -- do you have all the counts or only the proportions, for example? Do you know anything about the proportion of land held by neither farmers nor craftsmen? (to help get some sense of whether the dependence would matter much)
You could perhaps make an argument for using a Mann-Whitney, but you'd have to address several issues (for example, the discreteness - you can't just treat it as continuous, and the effect of the potential dependence).
If the counts are available, I'd be inclined to explicitly ignore the (probably mild) negative dependence and treat it as a set of comparisons of count proportions.
