A new study con­duct­ed by Professor Richard Berk of the UCLA Department of Statistics has iden­ti­fied sig­nif­i­cant sta­tis­ti­cal prob­lems with the data analy­sis used to sup­port recent stud­ies claim­ing to show that exe­cu­tions deter crime in the United States. In ​“New Claims about Executions and General Deterrence: Deja Vu All Over Again?,” Professor Berk address­es the prob­lem of ​“influ­ence,” which occurs when a very small and atyp­i­cal frac­tion of the avail­able data dom­i­nates the sta­tis­ti­cal results of a study. He found that this sta­tis­ti­cal prob­lem is found in a num­ber of recent stud­ies claim­ing to show that cap­i­tal pun­ish­ment deters vio­lent crime. The UCLA study con­duct­ed by Berk found that in many instances the num­ber of exe­cu­tions by state and year is the key explana­to­ry vari­able used by researchers, despite the fact that many states in most years exe­cute no one and few states in par­tic­u­lar years exe­cute more than five indi­vid­u­als. These val­ues rep­re­sent about 1% of the avail­able obser­va­tions that could have been used by researchers to draw con­clu­sions for ear­li­er stud­ies claim­ing to find that cap­i­tal pun­ish­ment is a deter­rent. In Professor Berk’s study, a re-analy­sis of the exist­ing data shows that claims of deter­rence are a sta­tis­ti­cal arti­fact of this anom­alous 1%. (Published on UCLA’s Web site, July 19, 2004). Read the study. See Deterrence and Resources.

Citation Guide