The most common approach to testing for the benefits of central-bank i
ndependence is argued to be methodologically flawed for the reason tha
t the measure of independence does not permit a test the hypothesis wh
ich suggests that these benefits exist. Particular doubt is cast on th
e merits of proposals to legislate to create independence. The argumen
t is elucidated by comparison with the testing of a hypothesis about t
he benefits of centralized and decentralized unions.