Adjusting the intervalling effect bias in beta : a test using Paris Bourse prices