Comparison of anonymous versus confidential survey procedures: effects on health indicators in Dutch adolescents.

INTRODUCTION

Self-report questionnaires have acquired an accepted role in youth research. They are easy to administer in a class-room setting, can address a wide range of topics that would otherwise require lengthy interviews, and usually require no complicated techniques of analysis. In general, the use of self-report instruments seems to eliminate variation attributable to interviewer characteristics (Tourangeau et al., 1997). Also, respondents are much more likely to disclose potentially embarrassing information when the questions are self-administered than when they are administered by an interviewer (Tourangeau et al., 2000).

With the increase of their use, the variety of methods of administration has increased. Some studies use questionnaires in a practically anonymous fashion, with no names on the questionnaire or questions within that could lead to identification. Other questionnaires are distributed with names and dates of birth on the cover, as they are part of a longitudinal design or serve as a preparation for later research. Mail surveys may be sent out to a specific person and returned anonymously. Whoever actually fills out the questionnaire and in what condition remains unclear. Recently, electronic questionnaire administration has added an extra dimension in the variety.

Although it could be expected that the method of assessment influences response, especially for questions on those kind of behaviors that seem to correlate negatively with social desirability, studies so far on this topic are not conclusive. In 1 of the earliest studies, Kulik et al. (1968) investigated whether anonymity is needed for a truthful self-report of antisocial behavior. A checklist of antisocial behaviors was filled out both anonymously and non-anonymously by high schools and delinquent boys. Both groups disclosed more antisocial behavior when protected by anonymity, but differences were comparatively small. Moreover, such differences were greater for minor violations (such as drinking alcohol to excess), than for serious ones (striking a teacher). They concluded that the importance of anonymity has been overemphasized in research on delinquency using self-report techniques. Similar studies were carried out on other subjects, mainly on substance use. King (1970) compared anonymous with precoded drug usage questionnaires in a mailed survey of undergraduates and found no statistically significant differences between the 2 groups with respect to number and percentage of returns, or reports of use of marijuana and LSD. Malvin and Moskowitz (1983) investigated differences between anonymous versus identifiable self-reports of drug attitudes, intentions, and use in junior high school students. Students in the identifiable condition reported less current use of cocaine than students in the anonymous condition. No effect was found for use of alcohol, barbiturates (tranquilizers), amphetamines (stimulants), marijuana (for girls), and cigarettes (for boys). The report of drug use and attitudes on drugs were not different in a study with a similar design by O'Malley et al. (2000). Bjarnason and Adalbjarnardottir (2000) found only small differences in a study with a corresponding design conducted among 10th graders.

Self-evidently, however, the sensitivity of a particular question is of importance. Thus, although Ong and Weiss (2000) found no differences in positive responses to questions about marijuana use, they did find differences to the questions about masturbation, cheating during exams, and stealing. The difference in response may be a result of cognitive factors such as the expectation that results will not be treated confidentially, despite the promise of the administering staff. Also, results may be influenced by the intention of the subject to depict a certain image of him--or herself--to the person who in the end sees the results. Intuitively, this social desirability would play a bigger part in confidential than in anonymous surveys.

In The Netherlands many municipal health services regularly use questionnaires on a variety of topics. A recent change in method in the Rotterdam preventive youth healthcare system raised questions about the implications of this change for epidemiological results: the outcome of individual questionnaires became part of the contents of an interview with the school nurse. This implied a change from anonymous self-report to a confidential procedure. As mentioned before, no consistent pattern appears from earlier studies. Therefore, the main research question is: Are there differences in the report of (psychological) health indicators between anonymous and confidential self-report questionnaires? Since there are substantial differences between boys and girls in health indicators in adolescence, we will study the influence of gender in this respect. Furthermore, the effect of social desirability, as 1 of the sources of response bias, will be taken into account in the analyses.

METHOD

Study Design

This study was conducted in 7 secondary schools in Rotterdam City, with all pupils in the 11th grade of general secondary education and preuniversity education …