See the page "Computation of Average Heterozygosity and Standard Error for dbSNP RefSNP Clusters" : http://www.ncbi.nlm.nih.gov/projects/SNP/Hetfreq.html
The data are indeed available for many different populations. SNP rs328 in LPL (lipoprotein lipase) has data from 23 different population-ss ID# combinations with regard to allele frequencies. The minor allele frequency ranges from 0.033 in YRI to 0.36 in a CEPH sample. Remember, ss IDs are folded into one rs ID.
Personally, I prefer to use data from HapMap when I am interested in genotypes of individual subjects. This allows us to calculate LD and such with tools in-house.