The purpose of this note is to show that the Mean Cost Rating (MCR), a
particularly useful statistic for measuring the accuracy of instrumen
ts used to predict recidivism, actually is a special case of Somers' D
. This latter statistic, normally used to measure concordance in a con
tingency table, is reported on the output of commonly used statistical
packages. Consequently, researchers who wish to use the MCR can obtai
n this statistic without any special programming or toilsome hand calc
ulation. However, some caution must be exercised in obtaining the MCR
from a statistical package because Somers' D has two forms, one with t
he rows as the independent variable and the other with the columns as
the independent variable. The MCR is the form of Somers' D that is opp
osite to what might be expected.