We give the large sample performance of the control quantile two sample tes
t relative to the two sample Wilcoxon rank sum test for various distributio
ns. We also give the exact null distribution in closed form for the case of
ties as well as for the untied case. Because of better efficiency for the
Cauchy distribution, and a simpler null distribution, the test seems an att
ractive substitute for the popular rank sum test for heavy tailed distribut
ions.