The Federal criminal sentencing guidelines struck down by the U.S. Supreme Court in 2005 required that males and females who commit the same crime and have the same prior criminal record be sentenced equally. Using data obtained from the United States Sentencing Commission’s records, we examine whether there exists any gender-based bias in criminal sentencing decisions. We treat months in prison as a censored variable in order to account for the frequent outcome of no prison time. Additionally, we control for the self-selection of the defendant into guilty pleas through use of an endogenous switching regression model. A new decomposition methodology is employed. Our results indicate that women receive more lenient sentences even after controlling for circumstances such as the severity of the offense and past criminal history.