Statistical tests of discrimination in punishment

A method, adopted from the labor econometrics literature, is proposed for detecting discrimination in punishment. The method requires the separate estimation of time served and punishment probability equations for, say, whites and blacks. The coefficients from the white equation are used to predict the punishment blacks would receive if treated like whites. A test of no discrimination against blacks is a test that the black punishment predicted by the black equation is equal to the punishment predicted by the equation using the white coefficients but the black endowments or characteristics. A further test is proposed that evaluates the economic efficiency of disparities in punishment. The test is restricted to measuring the recidivism effect of equality of treatment in punishment. The discrimination test and the efficiency test are illustrated using the U.S. Board of Parole data for 1972. Statistically significant racial disparities in punishment are uncovered and are found to be economically inefficient.