In Global Context: Mental Health Study
Indiana University . Bloomington . Indiana
Germany Sample Design
Survey Design and Methodology - Stigma Germany
Universe and Sample
The universe of this survey comprises all German speaking persons living in private households in the Federal Republic of Germany, at age 18 years and older.
In order to satisfy the scientific requirements of a random sample, the survey firm will employ the following, multi-layered sampling procedure according to the ADM-Masterplan.
First Sampling Layer
The survey firm is member of the "Arbeitsgemeinschaft Stichproben des ADM" (i.e. the representation of private market and social research agencies in Germany). The members of the ADM and MMC, Frankfurt/Main, own the ADM-Samplingsystem 2003. This system aims at the universe of private households and their respective household-members, living at the location of their main residency in all 16 Länder in the Federal Republic. The spatial grid of the system based on mixture of existing community borders and electronically-aided geographical delineations. As far as possible the geographical delineation is based on the smallest available administrative unit. Those areas are then separated further (using a geographical information retrieval system and the resident information) so that each unit contains on average 700 and a maximum of 1.400 addresses with minimized variance. In order to maintain a minimum size of 350 inhabitants per unit, some of the sampling units are merged into “synthetic” units.
This process produces 53.000 selection units which form the basis for drawing each sample.
Each sampling-unit is allocated a specific weight based on the number of all or all German private households. Since the number of private households is only available up to a specific administrative level, this number is re-calculated proportionally (i.e. to the number of persons eligible to vote or number of telephone households). Before drawing, the sampling-units are layered according to counties (Kreise) und BIK community-types (before: Boustedt). Within these layers, the sampling units are ordered according to their size, i.e. specific weights.
The essential principle of this system is to supply the involved institutes with samples that do not overlap and which can be easily cumulated. One of these ADM-nets consists of 210 sample-points in West Germany and 48 sample-points in East Germany. In order to construct a disproportional sample, the total number of sample-points can be increased by another 48 East German sample-points, or a completely new sample can be drawn which takes special requirements into account. The survey firm owns 16 of these nets exclusively.
For this project the survey firm will use one of its nets in West Germany (210 sample points) and one of its nets in East Germany (48 sample points). Thus, we can achieve on average approx. 3.8 net-interivews in West Germany, and approx. 4.2 net-interviews in East Germany per point.
According to the strict procedure of the ADM-sampling system, the interviews have to be conducted in exactly those sample-points which have been drawn. However, since there are not always interviewers available at every one of these sample-points, some of the sample-points may be substituted by others. The substituted sample-points will be within the same Land and be of the same community size as the original point. This will not affect the results obtained, yet reduce the costs for interviewing and travelling.
Second sampling layer
Within each sample-point, target households are selected through random route procedure. For each of the 258 sampling points one address is drawn at random. From this starting address, a cluster of max. 8 households is formed by selecting every third household by standard random route procedure.
Third sampling layer
Once the interviewer has reached the target-household, he has to identify the right target-person in those households with more than one potential target-person. This is done by listing all eligible household-members and identifying the target-person through a systematic key ("next-birthday-method"). If the interviewer does not reach this randomly selected target person, he must visit this household at least three times again to try to conduct the interview.
Fieldwork and Control
The survey firm will use approximately 200 interviewers for this survey. This will be interviewers who have in the past qualified especially well for face-to-face interviews. During the fieldwork period, the fieldwork management will continually keep contact with the interviewers. In addition to the written briefing, all interviewers can receive help by phone if needed.
The final data-set will document for every interview age, gender and level of education of the interviewer. We will be able to supply the information as requested in your draft for methodological report.
The survey firm will control 10% of all net-interviews. Controls are conducted via mail with the target-person and contain questions on the subject and date of the interview. If a mail contact cannot be achieved, controls are conducted – wherever possible – via telephone.
The survey firm has its own fieldwork management programme SAMSON, which processes all necessary information for fieldwork management and thus guarantees an optimal selection of interviewers and co-ordination of the fieldwork. The survey firm contracts approx. 2.400 face-to-face-interviewers, that are regularly briefed and controlled.
Transformation of household-sample into a person-sample
As only one person is interviewed per household, the sampling procedure described above yields a household sample, i.e. each household in the sampling universe has an equal chance to be sampled, but not every person. By transformation, the equal sampling probability of the households is replaced by the equal sampling probability of persons. For doing so, each case is multiplied with the number of persons, aged 18 years and older, living in the household. This number is then divided by the average number of persons living in households, in order to arrive at the real number of interviews. Since only one person per household will be interviewed, the described sampling-procedures will result in a household-sample, i.e. each household in the universe has the same probability to be included in the sample, but not each person. By transforming the sample, the equal chance for households will be replaced by the equal chance for persons. To do so, each case is multiplied with the number of household members age 18 and older. To obtain the actual number of cases, this number is divided by the average household size.
Alignment of the unweighted sample-structure to official statistics
Since losses in the sample are not equally spread over all population strata, the structure of the unweighted sample usually deviates systematically from official statistics. By means of weighting the sample structure is aligned mathematically with official statistics. Weighting criteria are age, sex, region (federal states) and political town size. The so gained three-digit weighting factor will be incorporated into the data file.
1022 E. Third Street, Bloomington, IN 47405 (812) 855-3841