And to estimate the share of prisoners in certain subgroups, it was necessary in some cases to use surveys conducted no more than once every five years, interpolating numbers in between and making the result fuzzier.
WSJ: Why Surveys Should Listen More to Prisoners