THE RESULTS OF THE GAO STUDY ON PROBATION AND RECIDIVISM INDICATED THAT PROBATION IS NOT REHABILITATING THE OFFENDER, NOR IS IT PROTECTING THE COMMUNITY. THE REPORT CLAIMED THAT ITS STATISTICAL ANALYSES DEMONSTRATED THE VALIDITY OF PREDICTION MODELS. THE RECOMMENDATIONS BASED ON THE STUDY WERE, ACCORDING TO THE AUTHORS, PREMATURE AT BEST. A REANALYSIS OF THE APPLICATION OF THE THREE 'BEST' MODELS (ACCORDING TO GAO) SUGGESTED THAT METHODOLOGICAL AND STATISTICAL ERRORS BIASED THE PRESENTATION AND RECOMMENDATIONS. THE THREE SUCCESSFUL MODELS WERE USED TO ANSWER THE FOLLOWING QUESTIONS: (1) WHO SHOULD RECEIVE PROBATION; (2) WHO WHOULD RECEIVE MINIMAL SUPERVISION WHILE ON PROBATION; AND (3) WHO SHOULD RECEIVE EARLY RELEASE. THE PREDICTIVE POWER OF EACH MODEL, AS ILLUSTRATED BY TABLES, WAS CLOSELY EXAMINED. THE MAIN CRITICISM OF THE GAO REPORT CONCLUSIONS WAS TWOFOLD: FIRST, THE GIVEN DATA SO FAR WOULD SUGGEST THAT THE IMPROVEMENT IN PREDICTION OBSERVED ON THE MODELS WAS EXAGGERATED; AND SECOND, THE CLAIM THAT THESE MODELS WERE EXTERNALLY VALID AND COULD BE EASILY APPLIED TO OTHER JURISDICTIONS DID NOT HOLD UP UNDER CLOSE EXAMINATION. ONE OF THE SHORTCOMINGS WAS IN THE BORROWING FROM OUTDATED, POTENTIALLY UNRELIABLE EQUATIONS; A PROBATION DEPARTMENT IS ENCOURAGED TO CONSTRUCT ITS OWN PREDICTIVE TOOL BASED ON LOCALLY AVAILABLE AND RELEVANT VARIABLES. IN CONCLUSION, THE AUTHORS DEMONSTRATE A NEW MODEL USING DEMOGRAPHIC VARIABLES FOR SUCCESSFUL AND UNSUCCESSFUL PROBATIONERS, PRESENTING METHODOLOGY AND RESULTS.