If there were reliable statistics, I'd much rather have the methods used to produce those statistics than the numbers.