The research findings present strong evidence, based on a specific data set, that the usual estimate of the deterrent impact of probation can be biased by the selection process involved in the probation program. This preliminary exploration of the data requires the use of bias-correcting mechanisms. This study involves the definition of selection and censoring biases, the demonstration of biases when the model is estimated by the least squares method, and the proposal of an alternative econometric model to correct these biases simultaneously. Results prove a clear reversal of the coefficients of interest as the biases are corrected, indicating that probation per se has no significant impact on drug offenders' duration of recidivism. 5 tables, 2 appendixes, 18 notes, and 40 references