>I believe that the justice department's major cross-sectional survey of
>crime rates, but the Bureau of Justice Statistics' flagship survey, the
>National Crime Victimization Survey (NCVS), is done through statistical
>sampling rather than through police reports. Rates of reporting (or
>conviction) thus aren't really germane to the question.>
The NCVS always shows there is much more crime than the DOJ's Uniform Crime Report which is basically just a tally of reported FBI index crimes. There are still a lot of problems with the NCVS though. For example: domestic violence is under-counted because the people doing the survey call and ask for the "head of household." When they ask if "you or anyone in your family" has been the victim of a violent crime, the male head of household is unlikely to respond that "Why yes, I beat the hell out of my wife 3 times so far this year."