There are two main clearinghouses for crime stats in this country, the UCR (The Uniform Crime Report) and the NCVS (National Crime Victimization Survey). The former is collected from police departments and thus only includes reported crime as recorded by the police. The latter is conducted by surveys and sampling and asks people (160,000 per year) if they were a victim of crime. They both can be useful in different situations, though I'm much more partial to the UCR.
Now here's the thing: The UCR says violent crime in 2012 is down 3% compared to 2010.
The NCVS says violent crime is up 39% in the past two years.
They can't both be right.
And I seriously suspect the NCVS is wrong.