The Credibility Effect and Rational Expectations: Implications of the Gramlich Study