Culture and auditor choice: A test of the secrecy hypothesis