Granted it is old data now, but I doubt that the overall figures have changed greatly. The study was carried out by the US Bureau of Justice Statistics and called "Recidivism of Sex Offenders released from Prison in 1994".
The following is from a blog,
Committee for Skeptical Inquiry
"According to a U.S. Bureau of Justice Statistics study (“Recidivism of Sex Offenders Released from Prison in 1994”), just five percent of sex offenders followed for three years after their release from prison in 1994 were arrested for another sex crime. A study released in 2003 by the Bureau of Justice Statistics found that within three years, 3.3 percent of the released child molesters were arrested again for committing another sex crime against a child. Three to five percent is hardly a high repeat offender rate.
In the largest and most comprehensive study ever done of prison recidivism, the Justice Department found that sex offenders were in fact less likely to reoffend than other criminals. The 2003 study of nearly 10,000 men convicted of rape, sexual assault, and child molestation found that sex offenders had a re-arrest rate 25 percent lower than for all other criminals. Part of the reason is that serial sex offenders—those who pose the greatest threat—rarely get released from prison, and the ones who do are unlikely to re-offend. If released sex offenders are in fact no more likely to re-offend than murderers or armed robbers, there seems little justification for the public’s fear and the monitoring laws targeting them. (Studies also suggest that sex offenders living near schools or playgrounds are no more likely to commit a sex crime than those living elsewhere.)
While the abduction, rape, and killing of children by strangers is very, very rare, such incidents receive a lot of media coverage, leading the public to overestimate how common these cases are. (See John Ruscio’s article “Risky Business: Vividness, Availability, and the Media Paradox” in the March/April 2000 Skeptical Inquirer.)"
As for murderers, a good article citing DoJ figures for inmates in California in 2002 is here at
Senator George Runner's blog.
"Consistent with other state and federal studies prisoners released following incarceration for murder were least likely to be charged with new crimes (40.7%) while prisoners released following a sentence for vehicle theft were most likely to be charged with new crimes (78.8%) within three years."