The Census Bureau provides frequency of all surnames occurring 100 or more times for the 2000 and 2010 census. Technical details of how the 2000 and 2010 data were collected can be found here (pdf) and here (pdf) respectively.
In the census data, for names with a count of 1--4, the counts are suppressed and replaced with the '(S)'. We replaced with '(S)' within a row by equally dividing the remaining percentage (100 minus the rest) across all the '(S).' For details, see the R script.