• No results found

Conclusion: still a consistent instrument

In 1987, when we carried out the first project aimed at measuring the prevalence and incidence of drug use, we already knew we would have to be patient for a while. We would have to wait until 1990 and beyond, to 1994, before we would we really be able to report something of value about drug-use behaviour. Any empirical ‘evidence’ resulting from measurements even if using such large- scale instruments as the survey applied here must be regarded a pale shadow of the real world. Any survey can ultimately be criticized for its moderate response, its instruction strategy, the approach, frame-errors, the way questions are asked, the people involved, and so forth. And although we again tried to operate as conscientiously and as carefully as possible to avoid all such criticism, in the end we agree with such a notion.

For that reason, already in 1987 we decided to pay attention to not only the quality of the instrument itself, but also to the consistency of the instrument in the long run. We therefore tried to keep the instrument as it was in former measurements. In short: a constant, unchanged, consistent instrument had to be applied. The instrument may be somewhat biased, but as long as it has not been changed - and assuming the bias is unchanged too - we at least can tell something about the changes going on.

Of course it is an illusion to think the instrument can really be kept unchanged in all its details. Sometimes a researcher, who played an important role in instructing interviewers, will no longer be able to join the project, or perhaps a new fieldwork organization has to be hired. On top of this, the interviewers who were involved in former years may not be the same as those who played a role in later years. Also specific events may have occurred, that may effect the results, and so on. We referred to some of these changes as possibly affecting the response rate. Overall, however, we assume these fluctuations will not have disturbed too greatly the consistency of the entire instrument applied. A comparison of the biases in the response-sample relation between 1990 and 1994 gives us an example of the consistency of the instrument, and also of its biases. In both 1990 and in 1994 the response-sample-population relations and deviations were most comparable. The only under-representation worth mentioning appeared to be of those who were born outside the Netherlands. The under-representation in 1994, however, was not different from that of 1990.

We can therefore conclude that the instrument is more or less constant, and that the biases too are probably constant. The changes in the use of drugs are therefore expected to be real changes, that must be ascribed to age cohort effects and other factors rather than to the instrument applied.

At this stage we think we should point out that only registered persons were included in the research. Tramps, street-persons, drug tourists and prisoners are not registered in the population registry, and therefore were not included in the survey.

11

Different Approaches

11.1 Introduction

It may be argued that it is important to investigate the effects of the application of other instruments. One reason for investigating the effects of (small) changes to the instrument applied, is technological progress that may help to obtain reliable response. More and more surveys are now making use of computer assisted methods. Interviews can easily be organized in such a way that response can be fed directly into a portable computer, thus achieving several gains in terms of reliability. A second reason to look at (slightly) different methods is that in surveys in which questions are asked that need some privacy in the interview situation in order to be answered correctly, the conditions should be created to guarantee that privacy. Here one might think of offering the interviewee the opportunity to administer the questionnaire him/herself. A third reason to look at different survey methods is ‘(future) international comparability’. Most surveys carried out in other countries have so far differed from those carried out in Amsterdam. The use of computers and the self-completion variation are ex- amples of such differences. To improve the comparison of the results of the Amsterdam survey with those of other cities, insight into the effects of these variations has to be improved.

There are reasons to expect effects from a variation in the way people are approached. One can think of the existence of some population categories comprising people who are still somewhat ‘afraid’ of computers (the elderly), or of differences in terms of the specific answers given depending on whether or not the questionnaire is self-completed. It can reasonably be expected that the use of illegal drugs will be mentioned more often if the interviewee’s privacy can be guaranteed. That expectation is supported by a recent publication of Aquilino (1994; see also Turner et al. 1992) on interview-mode effects in surveys of drug and alcohol use. He found a somewhat higher rate of admission of illicit drug use where interviewees were allowed to complete questionnaires themselves. He ascribed that effect to response anonymity. These effects were derived from analyses in the 37 largest Standard Metropolitan Statistical Areas in the United States. Only persons in the 18-45 age category were interviewed. Of that group, 25 per cent of those who came under the self-administered category stated that they had used cocaine at least once, whereas only 22 per cent of those who were personally interviewed admitted using it. However, American culture and the

different approaches

attitudes of Americans to the use of drugs may well differ from that of Dutch culture and the attitudes of the Dutch, particularly of those living in Amsterdam. Harrell (1985) provides evidence from which it can be concluded that even within cultures differences may show up as far as the response to questions are concerned. She found that differences in personal values, expectations, and reference group norms appeared to be key factors in how willing respondents are to provide authentic answers. Harrison (1995) too pointed at such effects on the validity of the methodology used. She argued that valid self-reporting of drug use is a function of the recency of the event, but also of the desirability of the drug, and nuances of the data collection methodology. Various strategies may, in other words, have various effects on different response categories in various circum- stances. It is important, therefore, that we too look at the effects of such variations in some detail, and that is the purpose of this chapter. We subdivided the sample into categories that had to be approached differently. Section 11.2 gives a brief comparison between those persons who participated in the so-called written version (the method we used in former years to measure drug use) and those who participated in the computer version. In Section 11.3 we focus attention on the difference between the interviewer-completed and the self-completed versions that were distinguished within the computer version approach. Section 11.4 summarizes the most relevant results.