A Note on Interviewer Effect in Self-Enumerative Procedures

Abstract

Most researchers are aware of the possibility of interviewer bias in personal interviews. In contrast, it is commonly assumed that data from self-enumerative procedures are free of such effects. Information collected in a study of the accuracy with which hospitalizations are reported, however, has led us to question the validity of this assumption. A sample of hospital records was selected, consisting of about 800 persons who had been hosptialized during the year preceding the period of data collection. The family names and addresses of these persons were assigned randomly to one of two data collection procedures; and within the procedures, assignments to interviewers were randomized. Both procedures began with a personal interview, using identical questions on demographic characteristics, illnesses, and other health events. In one procedure the interview ended with questions about hospitalizations for each family member during the preceding year. In the second procedure the questions about hospitalizations were omitted from the interview schedule and placed in a self-