Causes of mode effects: separating out interviewer and stimulus effects in comparisons of face-to-face and telephone surveys