INTRAVAL. Bureau voor sociaal-wetenschappelijk onderzoek en advies<
Between the Lines
Chapter 2   Methodology (part 1)
In this chapter we will discuss the way the research has been carried out. First, we will review some methods which are often used for the estimation of the extent of drug use. Then we consider snowball sampling and network analysis. After that attention will be paid to the research design of the cocaine study in Rotterdam. Finally the course of the research will be discussed.
2.1    Estimation methods in drug research
In the Netherlands and elsewhere a number of techniques have been applied in recent years to arrive at estimations of the extent of drug use (Hartnoll et al. 1985, Korf 1986, Intraval 1989 and 1991). All of these efforts referred to opiate addicts (including poly-drug users). Cocaine, however, is not an opiate. We are dealing here with a different category of hard drug users. The most important difference lies in the rapid and strong physical dependency developed by opiate users. This contributes to the fact that they come into contact earlier and more frequently with various drug programmes than cocaine users. Furthermore, the care and treatment centres can provide opiate users with methadon in the place of heroin, while no substitute exists for cocaine.
The clients' records kept by the centres for opiate users form a basis for a number of estimation methods. It must be mentioned in this context that the help given to opiate users is strongly institutionalised in the Netherlands. The opiate user with drug problems is well aware of which channels of assistance are open to him(1). Furthermore, other assistance agencies, doctors, the courts, police, family, friends and acquaintances also know where the opiate user can be referred to. These estimation methods are unsuitable, however, for determining the extent of cocaine use, because the users (as yet) rarely come into contact with care centres. In other words, these centres cannot function as a framework for estimation in the case of cocaine.
Police and court records form a frame of reference for a number of other estimation methods. However, while the police register opiate users as an entity, no records are kept of cocaine users as such. Rotterdam prison has kept such records since 1988, but naturally this applies only to detainees. Furthermore, four years of registration of cocaine users in the prison population is a relatively short time: the numbers registered will still be quite small.
Representative, or random, samples are used in many areas of research, also in surveys. A section of the population is selected at random in a sample. By the use of probability theory, the information derived from the sample can be translated to the population as a whole. A necessary condition to estimate the extent is that the item to be measured, for instance cocaine use, occurs frequently enough in the sample. Two other problems play a role in the question of the extent of use, namely non-response and reliability. Much depends on the probability that a user will admit to using cocaine. In the Netherlands the use of cocaine is not a criminal offence, but possession, transportation and manufacture of the drug is. Cocaine users therefore risk being arrested for possession of an illegal drug. It is to be expected that this will make a user reluctant to admit to using cocaine. This means that the survey method will probably provide an underestimation. On the other hand, some people still regard cocaine as a status drug. Therefore, some users will want to let it be known that they can afford the drug. What's more, cocaine has as yet not such a bad name as, for example, heroin.
A 1988 study of the population of Amsterdam shows that an estimation of the extent of drug use based on a survey could apply to Rotterdam only as far as the life-time prevalence (taken the drug at least once) and the last-year prevalence (having used the drug last year) of the category 17 to 49 years are concerned (Sandwijk et al. 1988). This method seems less suitable for persons who use the drug at least once a month, because the percentages will then almost certainly be too low. The extent to which the figures for Rotterdam differ from those for Amsterdam is relevant here. If the expected percentages are higher in Rotterdam, then the item to be measured will occur sufficiently frequently. However, a difference of minus one percent compared with the Amsterdam results would make an estimation of extent of use based on survey research a risky operation. The last-year prevalence in Rotterdam would then be less than 1%. It would take a very large (and therefore costly) survey to obtain accurate results over the size of such a small sub-population. From the above it can be concluded that a survey is not the appropriate method to study the nature and size of cocaine use.
Nomination technique
The nomination technique is an estimation method which has often been used in research among opiate users (Hartnoll et al. 1985, Korf 1986, Intraval 1990 and 1991). In this method, users are asked to name friends who use hard drugs, and whether these friends have been in touch with, for example, drug programmes in the last twelve months. On the basis of the ratio of programme to non-programme and figures from the drug assistance agencies an estimate can be made of the number of users. At the beginning of this chapter however it was already indicated that the drug programmes cannot function as a frame of reference for persons who only use cocaine. The drug assistance agencies in Rotterdam have only a small number of people who only use cocaine, while the police does not collect figures on cocaine users. This means that the method of nomination is either not applicable.
This method consists of comparing two independent samples from the same population. The number of people who appear in both samples is noted. On the basis of this overlap and the size of both samples it is possible to estimate the size of the whole population. If the overlap is great, the population is relatively small. If, on the other hand, the overlap is small, then the population must be relatively large. In practice however, especially for inaccessible groups such as drug users, it is extremely difficult to draw two independent samples from the population. A common solution is to use two separate existing data files. For opiate users this generally means files from the police and drug programmes. However, this method has the same problem as the nomination technique: drug assistance agencies and police have too few figures available.
2.2    Snowball sampling
The method of snowball sampling was originally developed for analyzing social structures within society. Already in 1958 Coleman introduced the snowball sample as a method of data collection which also takes into account the social environment. Up until now this method has been used in drug research mainly for contacting respondents. By using this method, however, it is also possible to gain insight into relationship structures in which cocaine plays a role.
The method of snowball sampling also provides the possibility of estimating the total number of users. The starting point is the initial sample of users, each of whom is asked to name all the other users he knows. The newly named users form the following 'layer' of the snowball and so it goes on. To ensure a statistically valid estimate, the following considerations are important.
The initial group of users of the snowball must resemble as closely as possible a random or stratified random sample of users. This could be achieved by, for example, first taking a random sample of residents of the municipality. This could be stratified so that groups in which a higher proportion of cocaine users is assumed are more highly represented. Those elements of the sample (residents of the city) who report that they use cocaine are then used as the initial group for the snowball. If the beginning of the snowball is not strictly random, there is the risk that the results will be biased. It would then be necessary to take into account the differences in probability that the various groups will be represented in the initial group.
In estimating the number of users by the snowball method, the capture-recapture method must be used: how many of the users named were already in the sample, and how many are genuine new names?
The size of the initial group is probably of great importance to the accuracy of the final estimate; if it is too small, meaningful results are not to be expected. The required size of the initial group would depend, among other things, on the network structure, that is the pattern of mutual acquaintance among users, and, of course, on the willingness of users to name other users(2).
Besides providing an estimate of the extent of use, the snowball method gives, just as other sampling methods, information about the composition of the user population. It should be noted here that the snowball sample is in itself not a representative group of users: better-known users have a much greater chance of being included in the sample than less well-known users. This bias will need to be taken into account when interpreting the data on the composition of the snowball sample.
2.3    Network analysis
Social network analysis is an approach to social phenomena in which the emphasis is placed on the relations between people (or other units of analysis, for example firms or countries), and in which graph theory is used (see for example Stokman 1982). Various sorts of relationships are possible. Relevant sorts of relationships in the study of cocaine use are, amongst others: A knows B and is aware that he uses cocaine; A started using cocaine under influence of (or after an introduction by) B; A buys cocaine from B. A network approach was already used in the nomination and snowball techniques without being named as such. The first type of relationship mentioned (acquaintance and knowledge of use) played a role here. Network theory is also used internationally for research into inaccessible populations or unsanctioned behaviour; see for example Sudman et al. (1988), Laumann et al. (1989) and Spreen (1992).
Besides being used in estimation methods, network analysis as such is also a useful approach for research into cocaine use. Using network-based sampling methods, information can be obtained on relations between users, which can in turn be analyzed using the methods of network theory. Fundamental questions are for example the degree to which users are aware of each other's cocaine use, the extent to which the user groups are divided into smaller sub-circuits, and the type of people who occupy the central position in such circuits. If the link between two circuits(3) of users is formed by a small number of people, it could be interesting to know what sort of people they are. Network information can also shed significant light on distribution patterns of cocaine use. For example: are a few individuals with a special role in the user world responsible for this distribution, or do all users share this distribution task more or less equally?
previous   next
Table of Contents
Chapter 1    Introduction
Chapter 2    Methodology
Chapter 3    General impressions
Chapter 4    The main characteristics
Chapter 5    Typology
Chapter 6    Spread, dispersion and extent
Chapter 7    Conclusions and discussion
Appendix A    Glossary
Appendix B    Occupation classification
Appendix C    Patterns of use
© INTRAVAL, Groningen-Rotterdam .
Deze site wordt onderhouden door De Poel Webdesign.