I present a model in which relative standing determines the level of achievement needed to pass a test. A candidate passes a test (or gains the credential) provided that his performance compares favorably to people who have passed the test recently. I argue that standards are likely to decline...